Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Empirical in-stream flow assessment tools for British Columbian channels McParland, Daniel 2013

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

Item Metadata

Download

Media
24-ubc_2013_fall_mcparland_daniel.pdf [ 2.6MB ]
Metadata
JSON: 24-1.0074274.json
JSON-LD: 24-1.0074274-ld.json
RDF/XML (Pretty): 24-1.0074274-rdf.xml
RDF/JSON: 24-1.0074274-rdf.json
Turtle: 24-1.0074274-turtle.txt
N-Triples: 24-1.0074274-rdf-ntriples.txt
Original Record: 24-1.0074274-source.json
Full Text
24-1.0074274-fulltext.txt
Citation
24-1.0074274.ris

Full Text

Empirical in-stream flow assessment tools for BritishColumbian channelsbyDaniel McParlandB.Sc, Queen?s University, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORALSTUDIES(Geography)The University Of British Columbia(Vancouver)September 2013c? Daniel McParland, 2013AbstractEmpirical hydraulic distribution equations have been proposed as simple and in-expensive alternatives to traditional data-intensive flow assessment methodologies.Two proposed depth and three proposed velocity empirical equations were com-pared to measured hdraulic distributions for two channels in the Interior Regionof British Columbia. Empirical velocity distributions adequately reproduced themeasured velocity distribution for both channels. An empirical depth distributionwas able to replicate measured depth distributions at a relatively undisturbed chan-nel (Harris Creek) but were unable to predict the measured depth distribution fol-lowing morphological change at a channel recently disturbed by forest fire (Fish-trap Creek). Furthermore, the empirical distributions were compared to modelleddepth and velocity distributions produced by a 2-dimensional hydrodynamic model(River2D). The empirical distributions provided reasonable representation of thehydraulic distributions for flows < 3 m3 s?1. At flows approaching bankfull theempirical methods, in particular the velocity equations, were unable to adequatelyreproduce the distributions produced in River2D. Additionally, a joint frequencydepth-velocity distribution was paired with habitat suitability indices to quantifyavailable habitat across a range of flows at Harris Creek using reach average hy-draulic conditions generated by River2D. The statistical habitat model producedsimilar habitat values to River2D at low flows and was able to recreate the generalshape and trends of the habitat indices. As well, a proposed at-a-station hydraulicgeometry simulator was used alongside a channel regime model to approximatereach average channel conditions at Harris Creek. The proposed hydraulic simula-tor was able to accurately predict reach average depth (mean error of 1.06%) andvelocity (4.47%) for discharges ranging from daily low flow to bankfull flow. Theiihydraulic simulator was coupled with the statistical habitat model to generate hy-draulic distributions and subsequently habitat indices for the modelled discharges.The incorporation of the regime models allows users to examine the influence ofvariable flow regimes and riparian vegetation (inherent of a changing climate) onavailable aquatic habitat. The proposed aquatic habitat model provides practition-ers with a low-input, user-friendly flow assessment tool that can be used for pre-liminary habitat assessments and basin-wide habitat studies in British Columbia.iiiPrefaceThis dissertation is original and unpublished work by the author, D. McParland.Assistance with developing the at-a-station hydraulic geometry simulator that isdescribed in Chapter 4 was provided by B. Eaton.ivTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xix1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Motivation for the study . . . . . . . . . . . . . . . . . . . . . . 11.2 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2.1 Physical methods . . . . . . . . . . . . . . . . . . . . . . 31.2.2 Habitat methods . . . . . . . . . . . . . . . . . . . . . . 41.2.3 Statistical methods . . . . . . . . . . . . . . . . . . . . . 61.3 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . 82 Study sites and data collection . . . . . . . . . . . . . . . . . . . . . 102.1 Harris Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10v2.1.1 Site description . . . . . . . . . . . . . . . . . . . . . . . 102.1.2 Data collection . . . . . . . . . . . . . . . . . . . . . . . 142.2 Fishtrap Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2.1 Site description . . . . . . . . . . . . . . . . . . . . . . . 182.2.2 Archived data . . . . . . . . . . . . . . . . . . . . . . . . 213 Empirical hydraulic distributions in British Columbian channels . . 223.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.2.1 Field data and 2-dimensional hydrodynamic simulations . 243.2.2 Empirical statistical distributions . . . . . . . . . . . . . . 253.2.3 Habitat indices . . . . . . . . . . . . . . . . . . . . . . . 273.2.4 Model evaluation . . . . . . . . . . . . . . . . . . . . . . 283.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 293.3.1 Measured data . . . . . . . . . . . . . . . . . . . . . . . 293.3.2 River2D data . . . . . . . . . . . . . . . . . . . . . . . . 373.3.3 Habitat model . . . . . . . . . . . . . . . . . . . . . . . . 413.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 463.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 494 Evaluation of a hydraulic geometry simulator in British Columbianchannels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.2 Rational regime model . . . . . . . . . . . . . . . . . . . . . . . 534.3 At-a-station hydraulic geometry simulator . . . . . . . . . . . . . 554.4 Application of a habitat model . . . . . . . . . . . . . . . . . . . 594.5 Model testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 604.6 Sensitivity analyses . . . . . . . . . . . . . . . . . . . . . . . . . 664.6.1 Harris Creek . . . . . . . . . . . . . . . . . . . . . . . . 664.6.2 Fishtrap Creek . . . . . . . . . . . . . . . . . . . . . . . 764.7 Model limitations . . . . . . . . . . . . . . . . . . . . . . . . . . 774.8 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81vi5 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 835.1 Empirical hydraulic distributions in British Columbian channels . 835.2 ASHGS aquatic habitat model . . . . . . . . . . . . . . . . . . . 845.3 The future of aquatic habitat modelling in British Columbia . . . . 86Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88A Habitat Suitability Indices . . . . . . . . . . . . . . . . . . . . . . . 95B Relative depth and velocity distributions - Harris Creek . . . . . . . 99C ASHGS habitat indices and sensitivity analyses . . . . . . . . . . . . 107viiList of TablesTable 2.1 Classification of cross-sections by morphological unit . . . . . 13Table 3.1 Index of Agreement (I) for proposed statistical depth distribu-tions at Harris Creek . . . . . . . . . . . . . . . . . . . . . . . 30Table 3.2 Index of Agreement (I) for proposed statistical velocity distri-butions at Harris Creek . . . . . . . . . . . . . . . . . . . . . 32Table 3.3 Index of Agreement (I) for proposed statistical velocity distri-butions at Fishtrap Creek . . . . . . . . . . . . . . . . . . . . 33Table 3.4 Index of Agreement (I) for proposed statistical high flow depthdistributions at Fishtrap Creek . . . . . . . . . . . . . . . . . . 34Table 3.5 Index of Agreement (I) for proposed statistical 2007 high flowdepth distributions for 2007 high flow conditions at Fishtrap Creek 34Table 3.6 Index of Agreement (I) for proposed statistical velocity distri-butions for 2007 high flow conditions at Fishtrap Creek . . . . 36Table 3.7 Index of Agreement (I) for proposed statistical depth distribu-tions for low flow condition sat Fishtrap Creek . . . . . . . . . 38Table 3.8 Index of Agreement (I) for proposed statistical depth distribu-tions and River2D depth data at Harris Creek . . . . . . . . . . 41Table 3.9 Index of Agreement (I) for proposed statistical velocity distri-butions and River2D velocity data at Harris Creek . . . . . . . 42Table 4.1 Modelled d? using River2D and ASHGS for a range of flows atHarris Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . 62viiiTable 4.2 Modelled u? using River2D and ASHGS for a range of flows atHarris Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . 63Table 4.3 Predicted changes in climatic variables at Harris Creek underthe Canadian Center for Climate Modelling and Analysis A2Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67Table 4.4 Predicted changes in climatic variables at Harris Creek underthe Canadian Center for Climate Modelling and Analysis B1Scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 68ixList of FiguresFigure 2.1 Location (inset) and boundaries of the Harris Creek watershed,tributaries, and study site . . . . . . . . . . . . . . . . . . . . 11Figure 2.2 Location of banks, bars, and thalweg, and placement of crosssections and depth loggers at Harris Creek . . . . . . . . . . . 12Figure 2.3 Longitudinal bed profile and water surface elevation along thethalweg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Figure 2.4 The position of the ADCP boat in the water as well as the co-ordinate system of the three emitted beams . . . . . . . . . . 16Figure 2.5 Grain size distributions determined from the Wolman pebblecount procedure at two bar heads, a riffle, and a pool . . . . . 17Figure 2.6 The recorded stage at data loggers located at the upstream anddownstream ends of the study reach. . . . . . . . . . . . . . . 18Figure 2.7 Location of Fishtrap Creek?s watershed, the extent of the for-est fire, and location of the study reach and Water Survey ofCanada stream gauge . . . . . . . . . . . . . . . . . . . . . . 19Figure 2.8 Location of banks and bars from 2006 to 2008 and placementof cross-sections at Fishtrap Creek . . . . . . . . . . . . . . . 20Figure 3.1 Relative depth distributions for low and high flow conditionsat Harris Creek. The bars represent the actual measured distri-butions. The lines are proposed statistical distributions. . . . . 30Figure 3.2 Relative velocity distributions for low and high flow conditionsat Harris Creek. The bars represent the actual measured distri-butions. The lines are proposed statistical distributions. . . . . 31xFigure 3.3 Relative velocity distributions for 2006 and 2007 high flowdata at Fishtrap Creek. The bars represent the actual measureddistributions. The lines are proposed statistical distributions. . 32Figure 3.4 Relative depth distributions for 2006 and 2007 high flow con-ditions at Fishtrap Creek . . . . . . . . . . . . . . . . . . . . 33Figure 3.5 Relative depth distributions for 2007 high flow conditions be-fore and after a rapid morphological change at Fishtrap Creek.The bars represent the actual measured distributions. The linesare proposed statistical distributions. . . . . . . . . . . . . . . 35Figure 3.6 Relative velocity distributions for 2007 high flow conditionsbefore and after a rapid morphological change at Fishtrap Creek.The bars represent the actual measured distributions. The linesare proposed statistical distributions. . . . . . . . . . . . . . . 36Figure 3.7 Relative depth distributions for 2005 through 2008 at low flowconditions at Fishtrap Creek. The bars represent the actualmeasured distributions. The lines are proposed statistical dis-tributions. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37Figure 3.8 Relative depth and velocity distributions for Harris Creek at Q= 0.75 m3 s?1. The bars represent distributions produced byRiver2D. The lines are proposed statistical distributions. . . . 38Figure 3.9 Relative depth and velocity distributions for Harris Creek at Q= 1.93 m3 s?1. The bars represent distributions produced byRiver2D. The lines are proposed statistical distributions. . . . 39Figure 3.10 Relative depth and velocity distributions for Harris Creek at Q= 5.00 m3 s?1. The bars represent distributions produced byRiver2D. The lines are proposed statistical distributions. . . . 39Figure 3.11 Relative depth and velocity distributions for Harris Creek at Q= 19.00 m3 s?1. The bars represent distributions produced byRiver2D. The lines are proposed statistical distributions. . . . 40Figure 3.12 WUA produced by River2D and a proposed joint frequencystatistical distribution model for adult rainbow trout at HarrisCreek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43xiFigure 3.13 WUA produced by River2D and a proposed joint frequencystatistical distribution model for juvenile rainbow trout at Har-ris Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Figure 3.14 WUA produced by River2D and a proposed joint frequencystatistical distribution model for spawning rainbow trout at Har-ris Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44Figure 3.15 WUA produced by River2D and a proposed joint frequencystatistical distribution model for adult smallmouth bass at Har-ris Creek . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Figure 3.16 WUA produced by River2D and a proposed joint frequencystatistical distribution model for adult longnose dace at HarrisCreek . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Figure 3.17 WUA produced by examining depth and velocity distributionsseparately for adult rainbow trout . . . . . . . . . . . . . . . 48Figure 4.1 The assumed channel geometry and characteristic rooting depth,H, embedded within UBCRM . . . . . . . . . . . . . . . . . 54Figure 4.2 Cumulative distribution functions of depth for b values of 0.2,0.5, and 0.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . 56Figure 4.3 Predicted reach average channel geometry at Harris Creek . . 61Figure 4.4 HV produced by River2D, Schweizer et al., and ASHGS foradult rainbow trout . . . . . . . . . . . . . . . . . . . . . . . 64Figure 4.5 HV produced by River2D, Schweizer et al., and ASHGS forjuvenile rainbow trout . . . . . . . . . . . . . . . . . . . . . 64Figure 4.6 HV produced by River2D, Schweizer et al., and ASHGS forspawning rainbow trout . . . . . . . . . . . . . . . . . . . . . 65Figure 4.7 HV produced by River2D, Schweizer et al., and ASHGS foradult smallmouth bass . . . . . . . . . . . . . . . . . . . . . 65Figure 4.8 HV produced by River2D, Schweizer et al., and ASHGS foradult longnose dace . . . . . . . . . . . . . . . . . . . . . . . 66Figure 4.9 Response of Wb, db, and b to changes in Qb at Harris Creek. . 69Figure 4.10 Response of Wb, db, and b to changes in H at Harris Creek. . . 70xiiFigure 4.11 Adult rainbow trout WUA and HV at Harris Creek across arange of potential Qb. . . . . . . . . . . . . . . . . . . . . . . 71Figure 4.12 Adult smallmouth bass WUA and HV at Harris Creek across arange of potential Qb. . . . . . . . . . . . . . . . . . . . . . . 71Figure 4.13 Adult longnose dace WUA and HV at Harris Creek across arange of potential Qb. . . . . . . . . . . . . . . . . . . . . . . 72Figure 4.14 Adult rainbow trout WUA and HV at Harris Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 73Figure 4.15 Adult smallmouth bass WUA and HV at Harris Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 73Figure 4.16 Adult longnose dace WUA and HV at Harris Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 74Figure 4.17 Sensitivity of adult rainbow trout weighted usable area (WUA)at minimum mean monthly flow (low flow) and mean annualflow as well as the maximum WUA for a range of Qb and H. . 74Figure 4.18 Sensitivity of adult smallmouth bass WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as themaximum WUA for a range of Qb and H. . . . . . . . . . . . 75Figure 4.19 Sensitivity of adult longnose dace WUA at minimum mean monthlyflow (low flow) and mean annual flow as well as the maximumWUA for a range of Qb and H. . . . . . . . . . . . . . . . . . 75Figure 4.20 Response of Wb, db, and b to changes in H at Fishtrap Creek. . 77Figure 4.21 Adult rainbow trout WUA and habitat value (HV) at FishtrapCreek across a range of potential H. . . . . . . . . . . . . . . 78Figure 4.22 Sensitivity of adult rainbow trout WUA at minimum mean monthlyflow (low flow) and mean annual flow as well as the maximumWUA for a range of H at Fishtrap Creek . . . . . . . . . . . . 78Figure 4.23 Adult longnose dace WUA and HV at Fishtrap Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 79Figure 4.24 Sensitivity of adult longnose dace WUA at minimum mean monthlyflow (low flow) and mean annual flow as well as the maximumWUA for a range of H at Fishtrap Creek. . . . . . . . . . . . . 79xiiiFigure A.1 United States Geological Survey (USGS) depth and velocityhabitat suitability indexes (HSI) for adult rainbow trout . . . . 96Figure A.2 USGS depth and velocity HSI for juvenile rainbow trout . . . . 96Figure A.3 USGS depth and velocity HSI for spawning rainbow trout . . . 97Figure A.4 USGS depth and velocity HSI for adult smallmouth bass . . . . 97Figure A.5 USGS depth and velocity HSI for adult longnose dace . . . . . 98Figure B.1 Relative depth and velocity distributions for Harris Creek at Q= 1.00 m3 s?1. . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure B.2 Relative depth and velocity distributions for Harris Creek at Q= 1.23 m3 s?1. . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure B.3 Relative depth and velocity distributions for Harris Creek at Q= 1.32 m3 s?1. . . . . . . . . . . . . . . . . . . . . . . . . . 101Figure B.4 Relative depth and velocity distributions for Harris Creek at Q= 1.61 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 101Figure B.5 Relative depth and velocity distributions for Harris Creek at Q= 2.29 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 102Figure B.6 Relative depth and velocity distributions for Harris Creek at Q= 2.61 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 102Figure B.7 Relative depth and velocity distributions for Harris Creek at Q= 3.08 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 103Figure B.8 Relative depth and velocity distributions for Harris Creek at Q= 3.51 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 103Figure B.9 Relative depth and velocity distributions for Harris Creek at Q= 3.98 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 104Figure B.10 Relative depth and velocity distributions for Harris Creek at Q= 4.47 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 104Figure B.11 Relative depth and velocity distributions for Harris Creek at atQ = 5.55 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . 105Figure B.12 Relative depth and velocity distributions for Harris Creek at Q= 7.50 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . . 105Figure B.13 Relative depth and velocity distributions for Harris Creek at Q= 10.00 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . 106xivFigure B.14 Relative depth and velocity distributions for Harris Creek at Q= 15.00 m3 s?1 . . . . . . . . . . . . . . . . . . . . . . . . . 106Figure C.1 Juvenile rainbow trout WUA and HV at Harris Creek across arange of potential Qb. . . . . . . . . . . . . . . . . . . . . . . 108Figure C.2 Spawning rainbow trout WUA and HV at Harris Creek across arange of potential Qb. . . . . . . . . . . . . . . . . . . . . . . 108Figure C.3 Juvenile rainbow trout WUA and HV at Harris Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 109Figure C.4 Spawning rainbow trout WUA and HV at Harris Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 109Figure C.5 Sensitivity of juvenile rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as themaximum WUA for a range of Qb and H. . . . . . . . . . . . 110Figure C.6 Sensitivity of spawning rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as themaximum WUA for a range of Qb and H. . . . . . . . . . . . 110Figure C.7 Juvenile rainbow trout WUA and HV at Fishtrap Creek across arange of potential H. . . . . . . . . . . . . . . . . . . . . . . 111Figure C.8 Sensitivity of juvenile rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as themaximum WUA for a range of H at Fishtrap Creek. . . . . . . 111Figure C.9 Spawning rainbow trout WUA and HV at Fishtrap Creek acrossa range of potential H. . . . . . . . . . . . . . . . . . . . . . 112Figure C.10 Sensitivity of spawning rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as themaximum WUA for a range of H at Fishtrap Creek. . . . . . . 112Figure C.11 Adult smallmouth bass WUA and HV at Fishtrap Creek acrossa range of potential H. . . . . . . . . . . . . . . . . . . . . . 113Figure C.12 Sensitivity of adult smallmouth bass WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as themaximum WUA for a range of H at Fishtrap Creek. . . . . . . 113xvList of Variablesb index of channel shapeC speed of sound (m3 s?1)d depth (m)d? reach average depth (m)db mean depth at bankfull discharge (m)dmax maximum water depth (m)Dn particle size associated with the nth percentile (mm)Fr reach average Froude numberg acceleration due to gravity (m sec?2)H representative rooting depth (m)Ho initial representative rooting depth (m)n number of binsPwetted channel wetted perimeter (m)Q discharge (m3 s?1)Qb bankfull discharge (m3 s?1)Qbo initial bankfull discharge (m3 s?1)R hydraulic radius (m)Res flow resistanceS energy slope, estimated from water surface slope (m m?1)s shape parameter for statistical velocity distributionSmix shape parameter for joint frequency distributiont shape parameter for statistical depth distributionv velocity (m s?1)v? reach average velocity (m s?1)xviW reach average width (m)Wb mean width at bankfull discharge (m)? mean of distribution? standard deviation of distributionxviiGlossaryADCP acoustic Doppler current profilerADV acoustic Doppler velocimeterASHGS at-a-station hydraulic geometry simulatorHSI habitat suitability indexesHV habitat valueIFIM Instream Flow Incremental MethodologyLW large woodPHABSIM Physical Habitat Simulation ModelUBCRM University of British Columbia Regime ModelUSGS United States Geological SurveyWUA weighted usable areaxviiiAcknowledgmentsFirst and foremost, I would like to thank my supervisor Dr. Brett Eaton. Brett?stechnical expertise, practical thinking, and sense of humour allowed this projectto reach its potential and made the experience worthwhile. I am also grateful tomy two other committee members, Dr. Jordan Rosenfeld and Dr. Andre Zimmer-man, for their advice and encouragement throughout this journey. Many thanksto Alistair Davis, Byeong Gyu Kim, Holly Buehler, Lesley Winterhalt, and SarahDavidson for their hard work and enthusiasm while completing field work for thisproject. Special thanks to Aaron Tamminga for being a constant source of knowl-edge and humour. My parents and sister have supported me on all my endeavorsincluding this one and I cannot thank them enough for that. Sara made the journey,through the challenges and achievements, a worthwhile experience and I am trulygrateful for all of her support. Funding for this project was provided by the PacificInstitute for Climate Solutions.xixChapter 1Introduction1.1 Motivation for the studyAcross the world, increased out-of-stream water demand has led to a convoluteddispute between river development and aquatic conservation [Tharme, 2003]. Thecomplexity of fluvial systems makes finding a balance between development andconservation an extremely difficult task. The unknown impacts of climate and landuse change on channel dynamics complicates the matter further [Conallin et al.,2010, Anderson et al., 2006]. Practitioners have and will continue to need scientifictools to aid them in establishing flows that withhold the ecological integrity ofthe channel as well as satisfy flow abstraction demand [Hardy, 1998, Saraeva andHardy, 2009a, Jowett, 1997, Hatfield and Bruce, 2000]. This need has led to thescience of flow assessment.Environmental flow assessment methodologies were first developed in the Pa-cific Northwest at the end of the 1940s. Starting in the 1970s, new environmentaland freshwater legislation as well as pressure from practitioners led to significantprogress in the development of flow assessment tools [Tharme, 2003]. For instance,in 1986, the Canadian Department of Fisheries and Oceans implemented its HabitatPolicy which required no net loss of productive capacity of channels. This policyhighlighted the need to continually develop and refine aquatic habitat assessmenttools for Canadian channels [Ahmadi-Nedushan et al., 2006]. The range and scopeof environmental flow assessment tools has broadened over the past 20 years due to1advances in computational technology [Gard, 2009, Jowett, 1997, Conallin et al.,2010].A multitude of in-stream flow methodologies have been developed and imple-mented across North America with varying degrees of resolution, complexity, andsubsequently success [Annear et al., 2004, Tharme, 2003, Jowett, 1997, Saraevaand Hardy, 2009b]. In the flow assessment community, there has been both effortsto refine and extend existing methodologies as well as develop new tools [Hardy,1998]. Each methodology has its own unique set of advantages and limitations.The most commonly applied methodology in North America is the Instream FlowIncremental Methodology (IFIM) and in particular the Physical Habitat SimulationModel (PHABSIM) component. PHABSIM links a hydraulic model based on cross-sectional data to habitat suitability indexes (HSI) to quantify habitat for a givenreach as a function of flow [Hardy, 1998, Jowett, 1997]. This methodology is con-sidered the most comprehensive and legally defensible but is heavily criticized forbeing technical, time consuming, and expensive and for ignoring ecological inter-actions [Hatfield and Bruce, 2000, Saraeva and Hardy, 2009b].Low-input, transparent alternatives to PHABSIM have and need to be continu-ally developed and refined in order to give practitioners practical in-stream flowassessment tools [Conallin et al., 2010]. Empirical hydraulic distributions havebeen proposed as a simple alternative to PHABSIM assessments [Lamouroux et al.,1995, Lamouroux, 1998, Schweizer et al., 2007]. The ability of these empiri-cal methodologies to approximate reach average hydraulic distributions and thusquantify physical habitat availability in British Columbian channels is relativelyunknown. Furthermore, the ability of empirical equations to model future physicalhabitat availability under different climate and land use change scenarios needs tobe assessed. The following section will provide a literature review of the relevantin-stream flow methodologies used in British Columbia. This will be followed bythe research objectives of this project.21.2 Literature review1.2.1 Physical methodsHistorical methodologies use archived discharge data usually in the form of monthlyor daily flow records to make flow recommendations [Tharme, 2003]. These meth-ods provide minimum flow values and flow duration within the historic range thatare needed to support aquatic life. They are considered to be effective and appropri-ate in the preliminary stages of waterway development. Historical methods requirelittle to no fieldwork. The most well known and used historical flow method isthe Tennant [1976] method. It is the second most used flow assessment tool in theUnited States of America [Reiser et al., 1989].Historical flow methods are simplistic and are readily used when there is anestablished flow record. However, due to their simplicity, and thus their numerousassumptions, they have many shortcomings. These methodologies are unable toquantify available habitat. As well, minimum flows vary amongst aquatic speciesand life stages [Hatfield and Bruce, 2000, Tharme, 2003]. Furthermore, histori-cal methodologies ignore limiting conditions such as cover availability, minimumdepth for fish passage, excess velocities, stream temperature, and food availability[Hogan and Church, 1989, Jowett, 1997].Moreover, applying the same minimum flows to different channels will havecontrasting results due to differences in channel geometry [Jowett, 1997]. Mini-mum flow values embedded in historical methodologies are often based on flowsrelative to the mean annual flow [Tharme, 2003]. However, bankfull flows set thechannel geometry and thus potential habitat. The ratio of bankfull flow to mean an-nual flow for nival channels are much greater than pluvial channels. Applying thesame minimum flow values (e.g. 10% of mean annual flow) to channels with dif-ferent flow regimes and therefore channel geometry will correspond to contrastingwater levels relative to bankfull channel conditions and thus differences in usablehabitat.Hydraulic geometry assessment tools are based on cross-sectional data andinclude both ?at-a-station? and ?downstream? methods [Leopold and Maddock,1953]. Cross-sections are appropriately placed to survey the variation in mor-3phological and hydraulic conditions within a reach. Flow requirements are of-ten determined from the observed wetted perimeter at each cross-section [Jowett,1997]. Hydraulic geometry methods provide only summary geomorphic and hy-drologic conditions of the channel and do not offer detail on local variability withina reach [Hogan and Church, 1989]. However, hydraulic geometry methods can as-sess whether the geometry conditions are approaching a threshold (e.g. minimumacceptable depth) and thus indicate the need for a more robust habitat assessment[Jowett, 1997].1.2.2 Habitat methodsHabitat methods (or combined hydraulic-habitat methods) are the most commonlyused in-stream flow methodology in the world, and in particular in North America[Conallin et al., 2010, Tharme, 2003]. Habitat methods divide the reach into cellsof a similar size, often in the form of closely spaced cross-sections. Within eachcell a deterministic hydraulic model is incorporated with univariate HSI to quantifyhabitat suitability. HSI define the biological requirements (usually for depth, ve-locity, and substrate) of an individual aquatic species and life stage [Jowett, 1997,Lamouroux et al., 1998, Guay et al., 2000]. The sum of the calculated habitat suit-ability of all the cells determines the potential habitable area of the reach. Habitatmethods are considered to be a reliable method of assessing available aquatic habi-tat because they quantify habitat for the entire reach and use a biological compo-nent [Jowett, 1997].Physical habitat simulation modelThe most well-known and used habitat method in North America is PHABSIM[Jowett, 1997, Tharme, 2003, Hatfield and Bruce, 2000, Parasiewicz and Walker,2007]. PHABSIM [Bovee, 1982] was developed in the 1970s by biological andphysical scientists at the U.S. Fish and Wildlife Service. It is considered to bethe most comprehensive and legally defensible methodology currently available[Hatfield and Bruce, 2000, Tharme, 2003, Lee Lamb et al., 2004, Hardy, 1998].There have been regional modifications of the original model (e.g. RHYHAB-SIM, CASIMIR, EVHA); however, the underlying principles of PHABSIM are4maintained [Parasiewicz and Walker, 2007]. The primary output of PHABSIM isweighted usable area (WUA) which is the amount of area in the channel deemedhabitable. PHABSIM can calculate WUA across a range of flows for target aquaticspecies and life stages.Hydrodynamic modelsOver the past 20 years, advances in computational technology has allowed the un-derlying principles of habitat methods/PHABSIM to be incorporated into 2-dimensionalhydrodynamic models (e.g. River2D, CCHEW2D, FEWMS-2DH). Hydrodynamicmodels are considered superior and seen as a potential replacement of traditionalPHABSIM methods because they predict depth and velocity both laterally and lon-gitudinally throughout a reach by explicitly using mechanistic processes such asconversion of mass and momentum [Leclerc et al., 1995, Gard, 2009]. Hydrody-namic models avoid the confusion of transect placement as all mesohabitats withinthe reach are modelled. As well, 2-dimensional models perform better in compli-cated channels because they account for local bed topography and roughness andthey can be used to assess hydraulic conditions at different discharges for an estab-lished channel boundary. Currently there are finite difference, finite volume, andfinite element hydrodynamic models commercially available [Steffler and Black-burn, 2002].Criticisms of habitat modelsDespite their widespread use, PHABSIM and 2-dimensional hydrodynamic modelsare highly criticized, most notably for being costly, complex, and time-consuming[Hatfield and Bruce, 2000, Armour and Taylor, 1991, Lamouroux and Capra, 2002,Saraeva and Hardy, 2009a,b, Lamouroux and Souchon, 2002, Lamouroux andJowett, 2005]. They require intensive site-specific field data collection which in-cludes numerous point depth and velocity measurements along geo-referencedcross-sections, substrate and cover classification, and a complete topographic sur-vey [Lamouroux and Jowett, 2005]. Data collection is followed by meticuloustechnical analyses of calculated hydraulic conditions, the use of simulation mod-els, and the application of HSI [Saraeva and Hardy, 2009b].5Other criticisms of PHABSIM and 2-dimensional models are that the results arenot monitored or verified and the underlying assumptions of the models are rarelytested [Hatfield and Bruce, 2000]. Also, the application of these methodologieshave been limited to case studies as they require extensive calibration [Lamourouxand Capra, 2002]. As well, PHABSIM output can only represent current channelconditions. It is unable to predict available habitat following geomorphic change[Bovee et al., 1998]. Moreover, HSI have long been scrutinized for not being trans-ferable between streams and ignoring ecological interactions [Hatfield and Bruce,2000, Hardy, 1998]. Lastly, limited resources, lack of environmental data, andneed for immediate assessment often exclude these methodologies as legitimatein-stream flow tools for many practitioners [Conallin et al., 2010].1.2.3 Statistical methodsSimple alternatives to PHABSIM and 2-dimensional hydrodynamic models havebeen sought in recent decades. Prediction of depth and velocity distributions usingempirical statistical models have become increasingly common. Early work on em-pirical statistical equations for the purposes of aquatic habitat modelling began inFrance [Lamouroux et al., 1995, Lamouroux, 1998] and was later adopted in NewZealand [Lamouroux and Jowett, 2005, Schweizer et al., 2007]. Statistical methodsrelate the form and shape of reach-scale depth and velocity distributions to easilyattainable predictor variables. The predictor variables are usually a reach averagehydrological or geomorphological condition (e.g. reach average Froude number orreach average relative roughness). Research has yielded both independent distribu-tion equations for depth and velocity [Lamouroux et al., 1995, Lamouroux, 1998]as well as a joint frequency depth-velocity distribution equation [Schweizer et al.,2007].Statistical methods are desirable because they do not require a complete bedtopography survey and have low computational costs compared to habitat methodsand 2-dimensional hydrodynamic models. Statistical methods are thought to be ofparticular use for preliminary assessment of habitat conditions as well as simulatingchanging flow conditions and channel boundaries [Schweizer et al., 2007]. Theyhave also been proposed as a rapid assessment tool when numerous channels need6evaluation [Saraeva and Hardy, 2009a]. As well, Schweizer et al. [2007] suggeststhat empirical statistical equations have some degree of universality.Aquatic habitat models created by pairing empirical hydraulic equations withHSI have produced comparable habitat indices to more data-intensive habitat mod-els in French and New Zealand channels [Lamouroux and Capra, 2002, Lamourouxand Jowett, 2005]. Statistical hydraulic distributions have recently been evaluatedin channels in the Pacific Northwest. Saraeva and Hardy [2009a] examined distri-bution equations developed in France [Lamouroux et al., 1995, Lamouroux, 1998]in the Nooksack River watershed in Washington State. Their results showed aslight manipulation of the original velocity distribution equation allowed for ade-quate prediction of available habitat in channels with a mean annual flow less than3.5 m3 s?1. The empirical equations performed poorly in large and irregularlyshaped channels.Furthermore, Rosenfeld et al. [2011] examined a joint frequency distributionequation developed in New Zealand [Schweizer et al., 2007] and locally derivedgamma distributions among contrasting habitat types in a small (approximate bank-full discharge of 0.6 m3 s?1) coastal stream. They found that gamma distributionsin conjunction with simple hydraulic geometry adequately represented reach scalehydraulic conditions in the trout-bearing stream. The joint frequency distributionperformed poorly suggesting that empirical frequency distributions cannot be eas-ily transferred to other biogeoclimatic regions without some form of local calibra-tion.While empirical statistical methods are an intriguing alternative to habitat mod-els and 2-dimensional modelling they do come with limitations. First and foremost,statistical methods do not provide any information on the spatial distribution of hy-draulic variables unlike PHABSIM and hydrodynamic models [Ahmadi-Nedushanet al., 2006, Lamouroux et al., 1995]. As well, statistical methods are best suitedfor channels with relatively natural morphology and flow regime [Parasiewicz andDunbar, 2001, Schweizer et al., 2007]. There has been little use of these method-ologies outside of the geographies in which they were developed and the universal-ity of the distribution equations is questionable [Schweizer et al., 2007, Rosenfeldet al., 2011]. Also, the mechanistic rationale behind the empirical distributionsand how the distributions evolve with changing flow and morphology is poorly7understood [Rosenfeld et al., 2011]. Furthermore, habitat models associated withempirical hydraulic equations are often designed for individual species and are notsuited for the analyses of multiple species [Ahmadi-Nedushan et al., 2006].1.3 Research objectivesA review of the relevant literature in Section 1.2 highlights the need for further de-velopment of low-input, user-friendly aquatic habitat models for British Columbianchannels. The overarching objective of this research project is to determine ifand to what extent empirical hydraulic equations can be used alongside chan-nel regime models to adequately predict changes in available aquatic habitat forBritish Columbian channels stemming from changes in flow regime and riparianvegetation dynamics. In order to meet this overarching objective there are threesub-objectives that have to be considered:1. The applicability of depth and velocity empirical distribution equations de-veloped in France [Lamouroux et al., 1995, Lamouroux, 1998] and NewZealand [Schweizer et al., 2007] need to be evaluated on British Columbianchannels. British Columbia has many unique biogeoclimatic regions and us-ing distribution equations developed on different continents may be inappro-priate. The ability of statistical methods to reproduce both measured velocityand depth distributions as well as velocity and depth distributions producedfrom 2-dimensional hydrodynamic model simulations will be evaluated.2. Empirical hydraulic equations need to be tested in both a disturbed andundisturbed watershed. The underlying hypothesis is that the statistical meth-ods should be able to adequately predict depth and velocity distributions foran undisturbed watershed, as the predictive equations were developed onchannels that had relatively natural flow regime and morphology. However,the ability of statistical methods to adequately predict depth and velocity dis-tributions in morphologically disturbed reaches has yet to be fully explored.3. Evaluate whether future channel dimensions resulting from climate and landuse change can be adequately modelled with a channel regime model. Fu-ture reach average hydraulic conditions predicted with a regime model will8be paired with applicable empirical hydraulic equations to determine futurehabitat indices for a species found in British Columbian channels.The remainder of this dissertation is organized as follows. Chapter 2 reviewsthe study sites and relevant data collection procedures. Chapter 3 examines theability of statistical distributions to reproduce measured and modelled (hydrody-namic model) depth and velocity distributions as well as WUA in British Columbianchannels. Chapter 4 proposes a low-input, user-friendly aquatic habitat model.Chapter 5 provides a summary of key conclusions and provides insight on futureaquatic habitat modelling research.9Chapter 2Study sites and data collection2.1 Harris Creek2.1.1 Site descriptionHarris Creek is a tributary of the Shuswap River located in the Interior Region ofBritish Columbia near the town of Lumby (Figure 2.1). The drainage area of thechannel is 220 km2 with approximately half of the catchment area above 1500 melevation on the Okanagan Plateau [Fletcher and Wolcott, 1991]. The terrain ischaracterized by gneisses and plateau basalts that are covered by a thin layer ofglacial drift [Day and Fletcher, 1989]. Mean annual flood is 19 m3 s?1 but thelargest recorded flood reached 35 m3 s?1. Discharge was recorded by a WaterSurvey of Canada stream gauge (station no. 08LC042) just downstream of thestudy reach. Floods are dominated by melting snowpack on the plateau in late Apriland early May [Fletcher and Wolcott, 1991, Hassan and Church, 2001]). Meanannual flow is approximately 1.5 m3 s?1. Unseasonably high spring precipitationin 2012 caused flooding to persist into late June.A 150 m study reach measured along the thalweg was used for this project(Figure 2.2). The reach was previously a study site for investigating bed load[Church and Hassan, 2002, Hassan and Church, 2001] and precious metals [Dayand Fletcher, 1989, Fletcher and Wolcott, 1991] transport. The average gradient ofthe reach is 0.011 m m?1 (Figure 2.3) and widths range from 10 to 20 m with an10Figure 2.1: Location (inset) and boundaries of the Harris Creek watershed,tributaries, and study site11Figure 2.2: Location of banks, bars, and thalweg, and placement of cross sec-tions and depth loggers at Harris Creekaverage width of approximately 15 m. The riparian vegetation is mature coniferoustrees and there are few pieces of large wood (LW) within the channel.The upper reach is characterized by a relatively straight series of glides andruns. The middle and lower reach are comprised of an alternating pool-riffle se-quence with the presence of two large bars [Montgomery and Buffington, 1997].Twelve cross-sections (1 through 6 and A through F) perpendicular to flow wereestablished along the reach and were distributed to evenly represent four morpho-logical units (Figure 2.2). Classification of each cross section into one of the fourmorphological units was strictly from visual inspection (Table 2.1). Furthermore,pressure transducers surrounded by PVC pipe were embedded in the substrate atthe upstream and downstream ends of the reach to record water stage every 15minutes. A transducer was also attached to a riparian tree to record barometricpressure.12Figure 2.3: Longitudinal bed profile and water surface elevation along thethalwegTable 2.1: The morphological unit, its defining characteristics, and the repre-sentative cross-sections from Harris CreekUnit Characteristics Cross SectionsPool 0 - 0.5% gradient, slow velocity, deep 3, E, 5Glide 0.5 - 1% gradient, relatively slow, undisturbed water surface 1, A, FRun 1 - 2% gradient, turbulent flow, shallow 2, B, CRiffle 1 - 3% gradient, fast current, protruding substrata D, 4, 6The bar heads are comprised of cobble-gravel surface and gravel and coarsesand subsurface. The bar tails are dominated by small pebbles with coarse sandin the voids. The wetted bed is well armoured with a median surface grain sizeranging from 64 mm (pools) to 76 mm (riffles) and a median subsurface grain sizeranging from 22 to 45 mm. The organized bed structure leads to small bed loadfluxes at Harris Creek [Hassan and Church, 2001].132.1.2 Data collectionHigh flowsHigh flow data were collected between April 29 and May 9th at discharges rangingfrom 1.61 to 4.82 m3 s?1. Depth and velocity data were collected at cross-sections1 through 6 on three separate occasions. Hydraulic measurements at each cross-section were taken as close to the bank as possible and then approximately every0.5 m across the channel until the far bank.A hand-held SonTek (firmware version 3.3, software version 2.20) acousticDoppler velocimeter (ADV) was used to collect both depth and velocity data fordepths less than approximately 0.3 m (i.e. near the banks). Water depths weredetermined from a measuring rod attached to the ADV. The ADV probe was placedat 60% water depth for 30 seconds. A transmitter within the probe generates soundconcentrated in a narrow beam at a frequency of 10 Hz. The emitted pulse travelsthrough the water column interacting with particulate matter (mostly sediment andbubbles) causing the sounds to be reflected in all directions. Two receivers aremounted on the probe so that they are receiving signals from water located 10 cmfrom the tip of the probe [SonTek, 2007]. The Doppler shift determines the velocityof the water:Velocity =fobservedfsource?C (2.1)where fobserved is the received frequency, fsource is the frequency of the emittedpulse (10 Hz), and C is the speed of sound (331 m s?1). The probe was pointeddirectly into the direction of the current, leading to negligible lateral velocity. Ve-locity measurements were averaged over 30 seconds at each vertical.A floating QLiner acoustic Doppler current profiler (ADCP) was used to mea-sure velocity at verticals that had water depths greater than approximately 0.3 m(i.e. close to the thalweg). The ADCP is ineffective in shallow waters due to a 20 cmblanking distance. Similar to the ADV, the ADCP uses the Doppler shift to quantifyvelocity. Three 1 Hz beams are emitted from a sensor located on the front of theboat (Figure 2.4). The Doppler shift was recorded every second for a duration ofat least 30 seconds at each vertical. Beam 3 was ignored as the turbulent water at14high flows often caused the front of the boat and thus beam 3 to be lifted out of thewater resulting in erroneous measurements. Beam 1 and 2 recorded the Dopplershift in 10 cm increments from the sensor (i.e. the first measurement was taken at30 cm depth, then 40 cm depth, 50 cm so on and so forth).The location of all ADV and ADCP measurements were geo-referenced with atotal station surveyor (Leica TPS800 Series) and reflective prism. The depth thatcorresponded to each ADCP velocity measurement was recorded as the differencein altitude of the boat minus the altitude of the bed (bathymetric survey wouldtake place at low flow). ADCP measurements recorded for depths greater than themeasured depth were dismissed at each vertical. The remaining ADCP velocitydata was corrected for the pitch of the boat at the time of the measurement. Thecorrected velocity data were averaged between beam 1 and 2 and then averagedover the depth of the vertical.Discharge at each cross-section was determined by assigning each vertical awidth half the distance to its neighbouring verticals. The discharge within eachvertical was determined by multiplying the area of the assigned rectangle (depthx width) by the corresponding velocity. The bins that extended from the bank tohalf the distance to the near bank verticals were assigned a velocity of zero. Thedischarge of the verticals were then summed for the cross-section to determine thetotal discharge [Corbett, 1962].Low flowsLow flow data were collected between July 18 and 23 at discharges ranging from0.76 to 2.02 m3 s?1. Data were collected at cross-sections 1 through 6 and A to Fon three separate occasions. Hydraulic measurements at each cross-section weretaken as close to the bank as possible and then subsequently every 0.5 m across thechannel until the far bank. All velocity and depth data were measured with an ADVas described above.A complete bathymetric survey was conducted from approximately 30 m up-stream of cross-section 1 to 50 m downstream of cross-section 6. A total stationsurveyor and reflective prism were used to record the northing, easting, and eleva-tion of all points relative to a set starting point. The bed was surveyed in a series15Figure 2.4: The position of the ADCP boat in the water as well as the coor-dinate system of the three emitted beams [QLiner, 2005]of cross-sections perpendicular to flow. An elevation was recorded on top of bothbanks at each cross-section. Points within the cross-section were spaced to allowfor a representative bed topography. Once a cross-section was complete a newcross-section was established approximately 2 to 3 m downstream of the previouscross-section and the process was repeated. 1942 elevation points were collected.Collection of bed elevation data along cross-sections 1 through 6 was thoroughto allow for accurate interpolation of depths corresponding to high flow ADCPlocations. However, the unforeseen flooding events in late June lead to a noticeablerestructuring of the channel. Thus, bed topographies recorded along the cross-sections 1 through 6 at low flow were probably different than the topographies ofthe cross-sections when the ADCP data were collected. This could have led to notonly erroneous depth data for high flow ADCP measurements but also inaccuratemean velocity data as the calculation of velocity was dependent on the depth of theindividual vertical.Furthermore, the location and elevation of the thalweg were recorded with atotal station surveyor and reflective prism. Depth and velocity data were recordedalong the thalweg using an ADV. The survey data were used to estimate the reachaverage energy gradient of the channel.Grain size distributions were examined at the head of the two large bars (repre-16Figure 2.5: Grain size distributions determined from the Wolman pebblecount procedure [Wolman, 1954] at two bar heads, a riffle, and a poolsentative of within channel substrate) as well as in a pool and in a riffle. Distribu-tions were quantified using the Wolman pebble count procedure [Wolman, 1954].Recorded grain size distributions (Figure 2.5) were very similar to distributionsrecorded in the late 1980s and early 1990s at Harris Creek [Church and Hassan,2002, Hassan and Church, 2001].The pressure transducers were removed from the channel on October 18, 2012.The barometric pressures were subtracted from the pressure data recorded by thein-channel transducers to determine water stage as a function of time (Figure 2.6).The hydrograph is unusual for Harris Creek as peak discharges were pushed wellinto June due to the unseasonably late spring floods. This resulted in the high flowsampling period (dashed vertical lines on Figure 2.6) not occurring at the highestflows. However, high flow data were collected above the mean annual dischargefor Harris Creek. Furthermore, the shift in the hydrograph led to low flow sampling(dotted vertical lines) to occur around the mean annual discharge, instead of belowit. Low flows appeared to occur in late August and September.17Figure 2.6: The recorded stage at data loggers located at the upstream anddownstream ends of the study reach. The dashed vertical lines repre-sents the high flow sampling period and the dotted lines represents thelow flow sampling period2.2 Fishtrap Creek2.2.1 Site descriptionFishtrap Creek is a tributary of the North Thompson River located in the InteriorRegion of British Columbia approximately 50 km north of Kamloops (Figure 2.7).The nival channel drains a watershed of 158 km2 with a mean annual peak flow ofabout 7.5 m3 s?1 [Eaton et al., 2010a,c]. Summer and autumn flows are typicallyless than 0.5 m3 s?1. The watershed contains thick deposits of glacial drift withthin and poorly developed soils. The region is characterized by short and mildwinters and hot and dry summers [Leach and Moore, 2010].In August of 2003, a high intensity forest fire burned 62% of the watershed (re-18Figure 2.7: Location of Fishtrap Creek?s watershed (dark grey shading), theextent of the forest fire (dark black line), and location of the study reachand Water Survey of Canada stream gauge [Eaton et al., 2010c]19Figure 2.8: Location of banks and bars from 2006 to 2008 and placement ofcross-sections (inset) at Fishtrap Creek [Eaton et al., 2010a]fer to Figure 2.7) and all the riparian vegetation. Since the fire, Fishtrap Creek hasbeen a site of ongoing hydrologic and geomorphic research [Eaton et al., 2010a,c,Leach and Moore, 2011, 2010, Petticrew et al., 2006, Phillips et al., 2009]. Aswell, there was a significant input of large wood (LW) into the channel followingthe fire. A large morphological change began to take place in late April 2007,which is attributed to a loss of bank strength from the decaying root system [Eatonet al., 2010c]. Figure 2.8 documents the alteration of the banks and bars from 2006through 2008.A 440 m long study reach, measured along the thalweg, was established justupstream of a Water Survey of Canada stream gauge (station no. 08LB024). Thereach alternates between riffle-pool and plane-bed morphologies [Montgomery andBuffington, 1997] with an increase in pool-rifle morphology since the fire. Thereare several LW jams throughout the reach. The mean bed gradient is approximately0.02 m m?1 and has a relatively coarse substrate (D50 = 55 mm, D84 = 128 mm).The average channel width of the reach is approximately 10 m [Eaton et al., 2010a].202.2.2 Archived dataHigh flow cross-sectional data were collected during the spring runoff in 2006 and2007 by Dr. Brett Eaton?s graduate students and research assistants. For 2006,there are depth and velocity data for cross-sections 1 through 5 (see inset of Figure2.8). These data were collected from April 29 to May 10 during flows rangingfrom 3.7 to 7.5 m3 s?1. An ADCP was used to collect velocity at each vertical asdescribed above. Depths were calculated by subtracting the recorded elevation ofthe bed from the observed water surface elevations at 0.3 m intervals across thechannel.For 2007, depth and velocity data were collected at cross-sections 1,2,3 and 5.These data were collected from April 9 to 27 during flows ranging from 1.95 to5.68 m3 s?1. In shallow sections of the channel, depth and velocity were measuredusing an ADV (i.e. near shore verticals). An ADCP was used to collect velocity andsubsequently determine depth at verticals closer to the thalweg.Channel topography was measured during low flow conditions at cross-sections1 through 11 from 2005 to 2008. The water surface elevation was recorded at eachcross section during the topographic survey. Point depths were calculated as thedifference between the elevation of the bed and the water surface at 0.3 m intervalsacross the channel. Reach average depth (d?) for each year was calculated as thesection-weighted mean of the point depth data. Reach average velocity (v?) for eachyear was determined using the following equation:v? =QW ? d?(2.2)where Q was the recorded daily discharge on the day of the sampling measured atthe gauge downstream of the study reach and W is the reach average width.21Chapter 3Empirical hydraulic distributionsin British Columbian channels3.1 IntroductionThe hydraulic habitat is a principal determinant of within-channel ecosystem struc-ture and function [Allan and Castillo, 2007]. Anthropogenic channel alterationand increased out-of-stream water use in channels throughout the world has ledto extensive alteration of channel hydraulics. Over the past quarter century, therehas been increased pressure to develop tools to help practitioners assess hydraulichabitat conditions required to protect aquatic species [Jowett, 1997, Hardy, 1998,Tharme, 2003].The most commonly used habitat assessment tool in North America is the In-stream Flow Incremental Methodology (IFIM) and in particular its Physical Habi-tat Simulation (PHABSIM) component [Annear et al., 2004, Conallin et al., 2010,Hatfield and Bruce, 2000, Lamouroux and Jowett, 2005]. PHABSIM combines ahydraulic model, usually in the form of cross-sectional data with habitat suitabilityindices (HSI) to quantify habitat at the reach scale. There are many regional vari-ations of PHABSIM but the underlying principles remain the same [Tharme, 2003].PHABSIM is considered the most thorough and legally defensible aquatic habitattool available to practitioners [Saraeva and Hardy, 2009a]. Furthermore, mecha-nistic 2-dimensional hydrodynamic models are increasingly being used to predict22depth and velocity patterns in rivers [Schweizer et al., 2007, Gard, 2009]. Similarto PHABSIM, the hydraulic conditions produced by the 2-dimensional models arecombined with HSI to quantify habitat.Despite their wide spread use, PHABSIM and 2-dimensional hydrodynamicmodels have received their fair share of criticism, most notably for being techni-cal, time consuming, and for ignoring ecological interactions [Hatfield and Bruce,2000, Saraeva and Hardy, 2009a,b, Schweizer et al., 2007]. Moreover, often thegoal of aquatic habitat assessment is prediction of future habitat conditions. PHABSIMand 2-dimensional hydrodynamic models can only deal with current channel bound-aries and cannot consider future channel geometry. Thus, practitioners cannot usethese methodologies to model future habitat availability [Schweizer et al., 2007].Beginning in the mid-1990s, simpler, statistical approaches have been devel-oped and refined for the purpose of predicting the reach-average distribution ofhydraulic conditions in rivers. These approaches predict the form and shape of thedepth and velocity distributions from easily obtainable predictor variables [Lam-ouroux et al., 1995, Lamouroux, 1998, Saraeva and Hardy, 2009a, Schweizer et al.,2007]. The predictor variables are most often reach average hydrological and/or ge-omorphic conditions. To date, statistical methods have been primarily developedin French and New Zealand riverscapes. Depth and velocity distribution equationshave been treated as both independent [Lamouroux et al., 1995, Lamouroux, 1998,Saraeva and Hardy, 2009a] and as a joint frequency distribution [Schweizer et al.,2007].Statistical methods are desirable because they do not require detailed data onchannel geometry and have low computational costs. As well, conceptually theyhighlight the primary controls of hydraulic habitat for many different channels[Schweizer et al., 2007]. Statistical methods can also be used to model the dis-tributions of depth and velocity for predicted future channel geometries by simplyestimating the future reach average channel conditions. Additionally, they providea rapid assessment tool of hydraulic conditions because of their low-input nature[Saraeva and Hardy, 2009a, Schweizer et al., 2007].The overarching objective of the research presented in this chapter was to de-termine if empirical hydraulic distributions developed in France and New Zealandcould estimate reach average hydraulic distributions in British Columbia channels.23The performance of statistical methods on both a relatively undisturbed (HarrisCreek) and disturbed (Fishtrap Creek) channel were evaluated. As well, depthand velocity distributions produced by a 2-dimensional hydrodynamic model andempirical statistical distributions were compared across a range of flows at HarrisCreek. Finally, a low-input, rapid assessment statistical habitat model is proposedfor British Columbia channels.3.2 Methods3.2.1 Field data and 2-dimensional hydrodynamic simulationsPoint depth and velocity data were collected at Harris Creek just outside of Lumby,British Columbia at both low and high flow conditions in 2012. Furthermore,archived point depth and velocity from high flow conditions at Fishtrap Creek for2006 and 2007 are available as well as low flow point depth data from 2005 through2008. For more information on the data collection and data refinement proceduresrefer to Chapter 2.River2D was chosen as the 2-dimensional hydrodynamic model to be used forthis project because it is customized for aquatic habitat evaluation studies [Stefflerand Blackburn, 2002]. The model is a depth-averaged finite element model thatsolves both the basic mass conservation equation and two horizontal componentsof momentum conservation. The bathymetric survey conducted at Harris Creekin July 2012 was used as the bed topography for the simulations. The no-flowboundary nodes were set as the first and last points (on top of banks) along eachcross-section used during the collection of bathymetric data. A curvilinear triangu-lated mesh comprised of 6651 nodes was created from the inputted bed topographyfile [Steffler and Blackburn, 2002, Gard, 2009].The first simulation was run using a discharge (Q) of 1.32 m3 s?1. Watersurface elevations at each of the cross-sections were known for this particularflow. The inflow and outflow water surface elevations were approximated as theseboundaries were beyond the reach where hydraulic data collection occurred (re-fer to Chapter 2). The downstream water surface elevation was adjusted until anodal convergence between time steps of less than 1 x 10?6 m was reached. To24calibrate River2D the bed roughness was adjusted until the difference between themeasured and simulated water surfaces elevations at cross-section 1 was negligibleand the water surface elevations at the rest of the cross-sections differed by lessthan 3 cm. Bed roughness was considered uniform for the channel because therewas limited data on the spatial variability of substrate when the simulations wererun. A roughness value of 0.37 m was used for all simulations.Upon calibration, 18 simulations were run at Harris Creek for Q ranging from0.75 to 19 m3 s?1. The accuracy of River2D was deemed sufficient and the pointdepth and velocities produced by the model were considered to be reasonable rep-resentations of the actual depth and velocity fields. Reach average depth (d?) andvelocity (v?) were determined by weighting each point depth and velocity by thearea of the corresponding cell and then dividing by the total wetted area.3.2.2 Empirical statistical distributionsIndependent velocity and depth distributionsMost of the preliminary work of developing independent velocity and depth dis-tributions was conducted on French channels. Lamouroux et al. [1995] developedan empirical equation for predicting the relative point velocity (v/v?) distributioncomprised of a decentered (exponential distribution) and centered (Gaussian dis-tribution) models, with the proportion of each model varying according to a shapeparameter (s):f (x = v/v?,s) = s(3.33exp(?x0.193)+0.117exp(?(x?2.441.73)2))+(1? s)(0.653exp(?(x?10.864)2)) (3.1)s =?0.150?0.252ln(Fr) (3.2)where Fr is the Froude number. The first term in Equation 3.1 represents thedecentered model and the second term represents the centered model.25Similarly, Lamouroux [1998] proposed modelling the relative depth (d/d?) dis-tribution as a mixture of Gaussian and exponential models, with the proportion ofeach model varying according to a shape parameter (t):f (x = d/d?, t) = t ? exp(?x) + (1? t) ?0.951exp(?(x?10.593)2)(3.3)t =?0.7ln(d?) (3.4)where the first term in Equation 3.3 represents the exponential model and the sec-ond term is the Gaussian model.Saraeva and Hardy [2009a] found that Equation 3.1 was unable to capture thefull range of velocity data collected from channels in the Nooksack River basin inWashington State. They proposed a slightly altered empirical velocity distributionequation, using the same shape parameter (s):f (x = v/v?,s) = s(3.33exp(?x0.693)+0.117exp(?(x?81.73)2))+(1? s)(0.653exp(?(x?1.10.664)2)) (3.5)Joint frequency distributionBuilding upon the work using independent empirical distributions, Schweizer et al.[2007] proposed a joint frequency depth-velocity distribution that utilizes a singleshape parameter (Smix). Similar to the independent empirical distributions, the rel-ative velocity and depth distributions were predicted using a combination of cen-tered and decentered models. However, Schweizer et al. [2007] used a log-normaldistribution as opposed to an exponential distribution to model the decentered por-tion of the hydraulic distributions. The predictive joint frequency equations are asfollows:26f (v/v?,Smix) = (1?Smix) ?Nu(?uN ,?uN)+(Smix) ?LNu(?uLN ,?uLN) (3.6)where ?uN = ?uLN = 1, ?uN = 0.52, ?uLN = 1.19, andf (d/d?,Smix) = (1?Smix) ?Nd(?dN ,?dN)+(Smix) ?LNd(?dLN ,?dLN) (3.7)where ?dN = ?dLN = 1, ?dN = 0.52, ?dLN = 1.09, andln(Smix1?Smix)=?4.72?2.84 ? ln(Fr) (3.8)All the empirical equations presented above were compared to both the col-lected hydraulic data at Harris and Fishtrap Creek as well as hydraulic data mod-elled in 18 River2D simulations. The empirical distributions were examined acrossa range of 0 < v/v? < 3 and 0 < d/d? < 3 using 30 equidistant bins.3.2.3 Habitat indicesWeighted usable area (WUA) were generated for three aquatic species: Oncorhynchusmykiss (rainbow trout), Micropterus dolomieu (smallmouth bass), and Rhinichthyscataractae (longnose dace). Adult, juvenile, and spawning rainbow trout life stageswere examined. These species were chosen as they all present unique and contrast-ing preferences for depth and velocity. Adult rainbow trout prefer shallow channelswith moderate flow. Juvenile rainbow trout prefer slightly deeper and slower cur-rents in comparison with adult rainbow trout. Spawning rainbow trout reside invery specific combinations of velocity (0.49 to 0.92 m s?1) and depth (0.21 to2.5 m). Smallmouth bass (adult) prefer deep water and slow currents (i.e. pools).Longnose dace (adult) are a benethic species that reside in fast and shallow current(i.e. riffles). HSI curves produced by the United States Geological Survey (USGS)were used for the analyses and can be found in Appendix A.For the River2D simulations, WUA was calculated by inputting the chosen HSIdata into the program. At each node, a depth and velocity preference was deter-mined. These preferences (varied between 0 to 1) were multiplied together and27then multiplied by the area of the associated cell to determine the WUA of eachcell. The total WUA was determined as the sum of the WUA of all the individualcells.For the statistical distributions, a preference was determined for each bin bycomparing the absolute bin value to HSI. Absolute bin values were determined bymultiplying the d/d? and v/v? distributions by d? or v? respectively. The total suitabilityof the velocity and depth distributions were determined as follows for n number ofbins:Velocity Suitability =n?i=1bin frequency ?bin suitabilityn?i=1bin frequency(3.9)Depth Suitability =n?i=1bin frequency ?bin suitabilityn?i=1bin frequency(3.10)where n is the number of bins. The total WUA produced by the statistical distribu-tions was calculated from the following equation:WUA = Wetted Surface Area ?Velocity Suitability ?Depth Suitability (3.11)The WUA calculated using the statistical distributions does not contain any spa-tial explicit detail and thus is not a ?true? WUA measure.3.2.4 Model evaluationAll statistical analyses were conducted using Matlab (version 7.9.0) statistical pro-gramme. For the purposes of the evaluation, depths and velocities predicted by theempirical statistical distributions were treated as ?Predicted? (P). The depths andvelocities measured in the field or modelled using River2D will be treated as ?Ob-served? (O). Model performance evaluation was based on literature presented byWillmott [1982] and Willmott et al. [1985]. The primary evaluation tool will be theindex of agreement (I), a dimensionless index ranging from 1 (perfect agreement28between the distributions) to 0 (no agreement):I = 1?[n(RMSD)2n?1(|P?i |+ |O?i|)2](3.12)where n is the sample size, P?i = Pi - O?, O?i = Oi - O?, O? is the mean observed value,and RMSD (root mean squared difference) is calculated as follows:RMSD =[1nn?1(Pi?Oi)2]1/2(3.13)For the purposes of this project, I values greater than 0.95 will be considered astrong fit, I values greater than 0.9 will be considered a reasonable fit, and I valuesless than 0.9 will be considered a poor fit.3.3 Results3.3.1 Measured dataHarris CreekThe empirical depth distribution proposed by Schweizer et al. [2007] provided asuperior fit to the measured depth data at both low and high flows (Figure 3.1). Atlow flow, the measured data was relatively normally distributed with a small tail(positively skewed). Both the normal distribution and tail were adequately pre-dicted by Schweizer et al. [2007] distribution. The Lamouroux [1998] distributioncontains an exponential component which greatly over predicted low depths andunder predicted mid to high depths.At high flow, the measured depth distribution was approximately normal (highFroude number) with a very small tail. Again, the Schweizer et al. [2007] distri-bution adequately modelled the observed depth distribution and was the superiorfit (Table 3.1). Lamouroux [1998] greatly over predicted low depths due to itsexponential component.The three proposed empirical velocity distributions provided a solid fit to the29Figure 3.1: Relative depth distributions for low and high flow conditions atHarris Creek. The bars represent the actual measured distributions. Thelines are proposed statistical distributions. The Lamouroux [1998] dis-tribution was developed in France and the Schweizer et al. [2007] dis-tribution in New Zealand.observed velocity data at both low and high flow (Figure 3.2 and Table 3.2). Atlow flow conditions, the measured velocity distributions had a peak at very lowvelocities (slow moving water on channel periphery and over bars) as well as apeak close to v?. The Lamouroux et al. [1995] and Saraeva and Hardy [2009a]distributions are able to model both peaks using a mixture of the exponential andnormal distributions. The Schweizer et al. [2007] distribution was unable to capturethe peak at very low velocities (key habitat for some species and life stages) butTable 3.1: Index of Agreement (I) for proposed statistical depth distributionsat Harris CreekFlow Lamouroux SchweizerHigh 0.879 0.968Low 0.748 0.93930Figure 3.2: Relative velocity distributions for low and high flow conditions atHarris Creek. The bars represent the actual measured distributions. Thelines are proposed statistical distributions. The Lamouroux et al. [1995]distribution was developed in France, the Schweizer et al. [2007] distri-bution in New Zealand, and the Saraeva and Hardy [2009a] distributionin Washington State.adequately recreated the remainder of the measured distribution.The measured velocity data at high flow had similar form and shape to the lowflow velocity distribution. However, at high flow conditions there were less lowvelocities and the distribution in more normally distributed (high Froude number).Again, the Lamouroux et al. [1995] and Saraeva and Hardy [2009a] distributionswere able to model both the small peak at low velocities and the peak that is nor-mally distributed about the mean. The Schweizer et al. [2007] distribution is unableto capture the peak at very low velocities using a log-normal decentered model.Fishtrap CreekHigh FlowThe observed high flow velocity distribution at Fishtrap Creek for both 2006 and2007 is relatively normally distributed (Figure 3.3). There does not exist a peak31Table 3.2: Index of Agreement (I) for proposed statistical velocity distribu-tions at Harris CreekFlow Lamouroux Schweizer SaraevaHigh 0.971 0.955 0.968Low 0.986 0.957 0.984Figure 3.3: Relative velocity distributions for 2006 and 2007 high flow dataat Fishtrap Creek. The bars represent the actual measured distributions.The lines are proposed statistical distributions.at a very low velocities as seen at Harris Creek. All three proposed velocity dis-tributions provided a strong fit to the measured data. The Schweizer et al. [2007]velocity distribution provides the strongest fit (Table 3.3) for both years as it wasable to predict the high densities for bins surrounding v?.For 2006 depth data, there exists both a peak at very low depths and a slightlylarger peak close to d? (Figure 3.4). The Schweizer et al. [2007] depth distributionis able to capture the larger peak surrounding d? but erroneously predicts low den-sities for very shallow depths. The Lamouroux [1998] depth distribution was ableto capture the peak at very low depths using an exponential model as well as ade-32Table 3.3: Index of Agreement (I) for proposed statistical velocity distribu-tions at Fishtrap CreekYear Lamouroux Schweizer Saraeva2006 0.959 0.974 0.9062007 0.962 0.979 0.947Figure 3.4: Relative depth distributions for 2006 and 2007 high flow condi-tions at Fishtrap Creek. The bars represent the actual measured distri-butions. The lines are proposed statistical distributions.quately modelling the peak surrounding d? making it the superior statistical depthdistribution for 2006 (Table 3.4).The measured depth distribution for 2007 is rather peculiar as there exists apeak before and after d?. This unusual distribution is most likely attributed to therapid widening that began to occur at Fishtrap Creek in late April 2007 (i.e. duringthe field season) due to the decay of the riparian root system [Eaton et al., 2010a,c].Both the Schweizer et al. [2007] and Lamouroux [1998] distributions provide apoor fit (Table 3.4) as their normal distribution component are distributed about d?.33Table 3.4: Index of Agreement (I) for proposed statistical high flow depthdistributions at Fishtrap CreekYear Lamouroux Schweizer2006 0.960 0.9372007 0.873 0.872Measured depth and velocity data for 2007 was divided into pre-morphologicalchange and post-morphological change datasets and compared to the proposed sta-tistical hydraulic distributions. Before the morphological change, there are still twopeaks in the measured depth data on either side of d? (Figure 3.5). However, both ofthe proposed depth distributions provide reasonable fits to the measured data (Table3.5). Following the morphological change there is a very large peak at low depthswhich is attributed to shallow depths that form as the channel widens and spreadsonto the channel periphery. As well, there are little depth data surrounding d? but anoticeable presence of high depth data. The proposed statistical depth distributionsdo a poor job of modelling the post-morphological depth distribution.Table 3.5: Index of Agreement (I) for proposed statistical 2007 high flowdepth distributions for 2007 high flow conditions at Fishtrap CreekCondition Lamouroux SchweizerBefore morphological change 0.913 0.924After morphological change 0.710 0.664Prior to the morphological change, the measured velocity distribution is rela-tively normally distributed with a slight positive skewness (Figure 3.6). All threestatistical velocity distributions are able to model the measured distribution. TheSchweizer et al. [2007] distribution provides the best fit as it is able to capturethe high densities surrounding v? (Table 3.6). Following morphological change, themeasured velocity distribution becomes negatively skewed and there is a peak at34Figure 3.5: Relative depth distributions for 2007 high flow conditions beforeand after a rapid morphological change at Fishtrap Creek. The barsrepresent the actual measured distributions. The lines are proposed sta-tistical distributions.velocities slightly greater than v?. The Schweizer et al. [2007] and Lamouroux et al.[1995] provide a reasonable fit to the measured data. Unlike the measured depthdistribution, the velocity distribution did not undergo a drastic change in shape andform following morphological change.35Table 3.6: Index of Agreement (I) for proposed statistical velocity distribu-tions for 2007 high flow conditions at Fishtrap CreekCondition Lamouroux Schweizer SaraevaBefore morphological change 0.944 0.962 0.929After morphological change 0.903 0.915 0.888Figure 3.6: Relative velocity distributions for 2007 high flow conditions be-fore and after a rapid morphological change at Fishtrap Creek. The barsrepresent the actual measured distributions. The lines are proposed sta-tistical distributions.Low FlowAt low flow conditions the Schweizer et al. [2007] depth distribution was able toadequately model the positively skewed measured depth distributions in 2005 and2006 (Figure 3.7). The Lamouroux [1998] distribution under predicted mid to highdepth for these years. The 2007 and 2008 depth data were collected after the rapidmorphological change. Similar to high flow conditions, the proposed statisticaldepth distributions were unable to model the measured depth distributions follow-36ing morphological change (Table 3.7).Figure 3.7: Relative depth distributions for 2005 through 2008 at low flowconditions at Fishtrap Creek. The bars represent the actual measureddistributions. The lines are proposed statistical distributions.3.3.2 River2D dataThere was a gradual shift in both the depth and velocity distributions producedby River2D for Harris Creek towards negatively skewed distributions as flow ap-proached bankfull (See Figures 3.8 through 3.11 and Appendix B). As flow in-creased, the depth distribution went from a relatively normal distribution withplatykurtic kurtosis and slight positive skewness to a distribution with leptokurtic37Table 3.7: Index of Agreement (I) for proposed statistical depth distributionsfor low flow conditions at Fishtrap CreekYear Lamouroux Schweizer2005 0.889 0.9292006 0.882 0.9212007 0.891 0.7942008 0.816 0.731Figure 3.8: Relative depth and velocity distributions for Harris Creek at Q =0.75 m3 s?1. The bars represent distributions produced by River2D. Thelines are proposed statistical distributions.kurtosis and negative skewness at high flow. At low flows, the velocity distributionhad a small peak at very low velocities and a larger peak at velocities slight greaterthan v?. As flows increased, the peak at low velocities lessened and the velocitydistribution became more leptokurtic and negatively skewed. As flows approachbankfull, the depth and velocity distributions had similar shape and form due toa decrease in relative roughness and flows becoming more homogeneous acrossmorphological units [Stewardson and McMahon, 2002].In terms of the empirical depth distributions, the Schweizer et al. [2007] pro-38Figure 3.9: Relative depth and velocity distributions for Harris Creek at Q =1.93 m3 s?1. The bars represent distributions produced by River2D. Thelines are proposed statistical distributions.Figure 3.10: Relative depth and velocity distributions for Harris Creek at Q= 5.00 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.39Figure 3.11: Relative depth and velocity distributions for Harris Creek at Q =19.00 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.vides a far superior fit to the River2D depth distribution across the range of sim-ulated flows (Table 3.8). Low flows are particularly relevant because they are theflow conditions experienced in a channel most of the year and are the limiting flowfor many species and life stages [Dakova et al., 2000]. At low flows the Lamouroux[1998] distribution greatly over predicted low depths and under predicted mid tohigh depths whereas the Schweizer et al. [2007] distribution was able to capturethe normally distributed nature of the River2D depth data. As flows increased theLamouroux [1998] performance improved and it?s I value approached that of theSchweizer et al. [2007] distribution. However, the Schweizer et al. [2007] distribu-tion was hands down the superior distribution at flows (< 3 m3 s?1) experienced atHarris Creek for the majority of the year.At low flows (< 2.29 m3 s?1) all three proposed statistical velocity distribu-tions provide reasonable fits to the velocity distributions simulated using River2D.However, the Saraeva and Hardy [2009a] distribution provides the best fit as itmodelled both the peak at very low velocities and the larger peak that occurs atvelocities slightly greater than v? (Figure 3.9). The Schweizer et al. [2007] distribu-40Table 3.8: Index of Agreement (I) for proposed statistical depth distributionsand River2D depth data at Harris CreekQ (m3 s?1) Lamouroux Schweizer0.75 0.635 0.9871.00 0.668 0.9881.23 0.702 0.9861.32 0.715 0.9861.61 0.731 0.9801.93 0.754 0.9822.29 0.776 0.9832.61 0.791 0.9793.08 0.803 0.9773.51 0.812 0.9733.98 0.822 0.9694.47 0.830 0.9655.00 0.838 0.9605.55 0.849 0.9597.50 0.874 0.95010.00 0.893 0.94215.00 0.915 0.92719.00 0.922 0.917tion is unable to model the peak at very low velocities because it does not have anexponential component. At Q = 2.61 m3 s?1, the Schweizer et al. [2007] becomesthe superior fit as the peak at the low velocities becomes small and the densitiesaround the v? becomes large. At flows greater than 3.08 m3 s?1 all three proposedvelocity distributions become poor as they are unable to model the high densitiessurrounding v?.3.3.3 Habitat modelThe Schweizer et al. [2007] depth distribution was chosen as the empirical depthdistribution for the proposed habitat model because it provided the best fit to boththe measured depth distributions and depth distributions modelled using River2D.Furthermore, the Schweizer et al. [2007] velocity distribution was chosen as the41Table 3.9: Index of Agreement (I) for proposed statistical velocity distribu-tions and River2D velocity data at Harris CreekQ (m3 s?1) Lamouroux Schweizer Saraeva0.75 0.948 0.953 0.9681.00 0.944 0.952 0.9651.23 0.940 0.949 0.9591.32 0.939 0.945 0.9571.61 0.933 0.943 0.9511.93 0.923 0.937 0.9402.29 0.911 0.928 0.9282.61 0.898 0.918 0.9153.08 0.881 0.904 0.8983.51 0.861 0.889 0.8793.98 0.851 0.880 0.8674.47 0.831 0.863 0.8475.00 0.818 0.851 0.8335.55 0.809 0.842 0.8227.50 0.775 0.810 0.78410.00 0.759 0.793 0.76315.00 0.742 0.775 0.73719.00 0.729 0.760 0.719empirical velocity distribution that will be used for the proposed statistical habitatmodel. Overall it provided the best fit to the River2D velocity data. It was also ableto adequately model measured velocity distributions at Harris Creek (although theother two distributions were slightly better) and provided the best fit to the FishtrapCreek velocity data. As well, using the Schweizer et al. [2007] velocity distributionalongside the Schweizer et al. [2007] depth distribution requires the calculation ofonly one shape parameter and allows for consistency in the model.For all species/life stages under investigation there was a clear difference inabsolute values of WUA produced by River2D (i.e. a reasonable representation ofactual conditions) and the proposed statistical habitat model (Figures 3.12 through3.16). This was especially true at high flows as the ability of the empirical statisticaldistributions to model hydraulic conditions, in particular the velocity distribution,42became poor. For all species there is an over prediction of WUA at flows greaterthan 3.51 m3 s?1. However, for the most part the relative shape and trends of theWUA data as a function of discharge were similar between the proposed statisticalhabitat model and River2D. The location of peak WUA were slightly different forsome species, in particular spawning rainbow trout.Figure 3.12: WUA produced by River2D and a proposed joint frequency sta-tistical distribution model for adult rainbow trout at Harris Creek43Figure 3.13: WUA produced by River2D and a proposed joint frequency sta-tistical distribution model for juvenile rainbow trout at Harris CreekFigure 3.14: WUA produced by River2D and a proposed joint frequency sta-tistical distribution model for spawning rainbow trout at Harris Creek44Figure 3.15: WUA produced by River2D and a proposed joint frequency sta-tistical distribution model for adult smallmouth bass at Harris CreekFigure 3.16: WUA produced by River2D and a proposed joint frequency sta-tistical distribution model for adult longnose dace at Harris Creek453.4 DiscussionAt Harris Creek, a combination of exponential and normal models provided thebest fit to the measured velocity data as these distributions were able to adequatelycapture the peak at very low velocities (Figure 3.2). This could suggest that a sta-tistical distribution with an exponential and normal component provides a superiorfit to velocity data or that the log-normal model embedded in the Schweizer et al.[2007] velocity distribution needs refining. At Fishtrap Creek, there was not a dis-tinct peak at very low velocities and all three proposed statistical distributions wereable to capture the shape and form of the measured distribution.An empirical depth equations comprised of an exponential and normal model[Lamouroux, 1998] was unable to replicate the measured depth distributions atHarris and Fishtrap Creek. The frequency of low depths was considerably overpredicted using an exponential model. Saraeva and Hardy [2009a] found that theLamouroux [1998] depth distributions resembled actual depth conditions on chan-nels much smaller than Harris Creek and Fishtrap Creek but over predicted lowdepths on channels of similar size to Harris Creek and Fishtrap Creek. Statisti-cal distributions comprised of log-normal and normal components provide the bestrepresentation of actual depth conditions.The proposed statistical velocity distributions provide strong fits to the mea-sured distributions at both Harris and Fishtrap Creek. This suggests that the pro-posed statistical velocity distributions can adequately predict velocity distributionsfor both undisturbed and disturbed channels. Furthermore, the Schweizer et al.[2007] depth distribution provided a strong fit to the measured depth distributionsat Harris Creek as well as an adequate fit at Fishtrap Creek before morphologicalchange. The statistical depth distribution performed poorly at Fishtrap Creek fol-lowing morphological change at both high and low flow suggesting the Schweizeret al. [2007] depth distribution is unable to capture the unique depth distribution ina morphologically disturbed channel. Some of the channels in which the statisticaldepth distributions were developed had some form of flow regulation for flood con-trol or hydro power purposes. However, it does not appear that any of the channelsused to develop the statistical models were severely disturbed [Schweizer et al.,2007, Lamouroux et al., 1995, Lamouroux, 1998, Saraeva and Hardy, 2009a].46In general, as discharge increased the ability of statistical distributions to repro-duce hydraulic conditions simulated using River2D became poorer. The statisticaldistributions become questionable at discharges > 3 m3 s?1. The poor performanceof the empirical distributions at higher discharges stem from the statistical distribu-tions being developed on channels that were usually experiencing flows less thanmean annual flow (flows > 3 m3 s?1 are higher than mean annual flow at HarrisCreek). For instance, Schweizer et al. [2007] collected and modelled hydraulicdata for channels that were experiencing discharges 5 to 100 % of their mean an-nual flow. They warned that the use of statistical distributions at discharges greaterthan mean annual flow could lead to poor fits as relative roughness and channelheterogeneity decrease.Furthermore, Saraeva and Hardy [2009a] found that the statistical distributionswere only effective for channels with mean annual flow less than 3.5 m3 s?1 inthe Nooksack River basin in Washington State. This appears to coincide with thestatistical distributions performance at Harris Creek. It is encouraging that thestatistical distributions are adequately modelling depth and velocity distributionsat low flows. Low flows are most often the limiting flows for aquatic species andthus being able to model the distribution of these two important hydraulic variablesat low flows is critical [Hatfield and Bruce, 2000, Dakova et al., 2000].The proposed statistical habitat model is the early makings of a rapid habitatassessment tool for practitioners. The general shape and trends of the WUA curvecan be predicted using empirical statistical distributions. The errors in absoluteWUA values are of similar magnitude to errors observed by Saraeva and Hardy[2009a] using a comparable statistical habitat model. Differences in absolute WUAvalues were expected because of the different methodologies used by River2D andthe proposed statistical habitat model to calculate WUA. River2D calculates WUAusing a bivariate pair of depth and velocity at each node whereas the statisticalmethods calculate WUA by examining the depth and velocity distributions inde-pendently (i.e. not a true WUA calculation).Furthermore, errors in WUA arise from the inability of statistical velocity distri-butions to model actual conditions. This becomes apparent when calculating the in-dependent WUA of each hydraulic variable and comparing them between River2Dand the statistical habitat methods (Figure 3.17). The differences in depth WUA47Figure 3.17: WUA produced by examining depth and velocity distributionsseparately for adult rainbow troutproduced by the two models for adult rainbow trout is negligible. Velocity WUAis comparable between River2D and the proposed statistical habitat model at lowflow but significantly deviates at flows greater than approximately 3.5 m3 s?1. Aswell, biological models are more sensitive to velocity errors as velocity preferencesare usually more complex and dynamic (i.e. velocity HSI curves have many abruptshifts) in relation to depth preferences [Lamouroux, 1998]. Therefore, to improvethe proposed statistical habitat model, especially at high flows, statistical velocitydistributions need to be improved.There exists some obvious limitations to the statistical habitat model. First,the habitat model was developed and tested on one channel. Ideally, the statisticalhabitat model will be tested and refined on many British Columbian channels withvarying morphology and flows regimes. Furthermore, evaluation of the proposedhabitat model on more species and life stages is imperative. Saraeva and Hardy[2009a] found that a similar statistical habitat model adequately predicted habitatindices for adult and spawning fish but provided poor prediction for juvenile andfry species that prefer lower velocities.Moreover, cover and substrate preferences were ignored in River2D simula-48tions. Thus, WUA was calculated from only depth and velocity data in River2D,which made for a simple comparison between models. There is no simple way toincorporate cover and substrate data into the statistical habitat model which makesthe proposed habitat model appropriate for preliminary assessments but not for de-tailed analyses. Cover and substrate are extremely important to aquatic habitat,especially for younger life stages and spawning species [Bovee, 1982]. ComparingWUA that does not incorporate cover and substrate data for species that are verysensitive to cover and substrate conditions is dubious [Saraeva and Hardy, 2009a].As well, treating the hydraulic distributions produced by River2D as reasonablerepresentations of the actual depth and velocity fields is a valid assumption butan assumption nonetheless. There is certainly potential for River2D to produceslightly erroneous depth and velocity data and thus WUA.3.5 ConclusionEmpirical statistical distributions can be used for preliminary assessment of hy-draulic conditions in British Columbian channels. In particular, a joint frequencydepth-velocity distribution developed on New Zealand channels can reproduceboth measured depth and velocity distributions for an undisturbed channel. Thejoint frequency distribution was also able to reproduce depth and velocity distri-butions produced using a 2-dimensional hydrodynamic model (River2D) for flowsclose to and less than mean annual flow. As flows increase towards bankfull condi-tions prediction of the hydraulic distributions produced by River2D become poor.A statistical depth distribution containing a log-normal and a normal model wasable to replicate depth distributions at Harris Creek and at Fishtrap Creek priorto rapid channel widening in late April 2007. The empirical depth distributionsperformed poorly at Fishtrap Creek after the rapid morphological change, suggest-ing empirical distributions are unable to reproduce hydraulic conditions in recentlydisturbed channels.The early makings of a low-input, rapid aquatic assessment tool that can beused by practitioners across British Columbia for preliminary habitat assessmentsand basin-wide aquatic habitat studies is presented above. The model inputs aresimply reach average depth and velocity which can often be collected in one day49of field work. The statistical habitat model can highlight channels that are in needof further in-depth assessment as well as set the experimental boundaries for fu-ture aquatic habitat research [Hatfield and Bruce, 2000]. The general form andshape of WUA data produced by 18 River2D (data-intensive hydrodynamic model)simulations at Harris Creek were reproduced using the proposed statistical habitatmodel.The proposed aquatic habitat model needs further refinement. The ability ofempirical statistical distributions to replicate depth and velocity distributions inchannels that have different morphologies and flow regimes than Harris Creekneeds to be examined. The empirical velocity equations embedded in the statisticalhabitat model needs to be refined or replaced to allow for more accurate predictionof habitat indices. As well, the proposed statistical habitat model does not accountfor many environmental factors that determine habitat (e.g. cover availability, sub-strate). Finally, the proposed habitat model should be compared to River2D orPHABSIM outputs for more species and life stages. Using the proposed methodol-ogy as anything but a preliminary assessment tool is ill-advised.50Chapter 4Evaluation of a hydraulicgeometry simulator in BritishColumbian channels4.1 IntroductionIncreased flow abstraction for agriculture, industry, electricity production, and recre-ation has led to decreased flows in channels throughout the world. Practitioners arecontinually faced with the difficult task of recommending and enforcing in-streamflow requirements that can both maintain the ecological integrity of the channeland meet out-of-stream demand [Saraeva and Hardy, 2009a]. Changes in mag-nitude and timing of channel inputs associated with climate and land use changefurther complicates the task [Tharme, 2003].Considerable effort has been made in recent decades to improve in-stream flowassessment tools [Hardy, 1998]. The most well known and used aquatic habi-tat tool is the Instream Flow Incremental Methodology (IFIM) and in particularits Physical Habitat Simulation (PHABSIM) component [Lamouroux and Jowett,2005, Jowett, 1997]. PHABSIM [Bovee, 1982] combines point depth, velocity, andsubstrate measurements with habitat suitability indices (HSI) to calculate weightedusable area (WUA) for the reach under investigation [Saraeva and Hardy, 2009a].51Furthermore, the underlying principles of PHABSIM are being incorporated into2-dimensional hydrodynamic models as they can predict depth and velocity bothlongitudinally and laterally throughout a reach. Similarly, depth and velocitiesproduced using 2-dimensional models are combined with HSI to predict WUA. 2-dimensional models are desirable because they avoid problems with transect place-ment, they can model complex habitats (provided the underlying assumptions ofthe model are met), they take into account roughness and bed topography, and theyrely on mechanistic processes [Gard, 2009].Despite the widespread use of PHABSIM, and more recently 2-dimensional hy-drodynamic models, they do come with their fair share of criticism. The mostcommon criticisms are the expensive, time consuming, and technical nature ofthe models as well as their inability to model ecological interactions [Gard, 2009,Hatfield and Bruce, 2000, Hardy, 1998, Saraeva and Hardy, 2009a,b, Lamourouxand Souchon, 2002, Lamouroux and Jowett, 2005]. The models require intensivesite specific data collection including numerous point depth and velocity measure-ments, substrate and vegetation cover quantification, and complete bathymetricsurveys (essential for 2-dimensional modelling). Upon data collection, expertiseis needed in processing the data, calibrating the models, and applying appropriateHSI.Moreover, often the purpose of aquatic habitat modelling is the predictionof future habitat conditions. Quantifying the impact of proposed hydroelectricdams, water diversions, rehabilitation projects, and climate and land use changeon aquatic habitat all require prediction of future channel dimensions. PHABSIMand 2-dimensional models are adequate at predicting current WUA using measuredchannel conditions [Saraeva and Hardy, 2009b, Gard, 2009]. However, as the flu-vial system is perturbed there are inherent changes in stream morphology and sub-strate (i.e. changes in boundaries) which greatly influence physical habitat. Theboundaries of PHABSIM and 2-dimensional models cannot be readily changed thusthese tools cannot be used to predict future hydraulic conditions and subsequentlyavailable habitat [Hatfield and Bruce, 2000].There has been a push to develop more simplistic and cost-effective tools forwater practitioners that can be readily applied to multiple channels in a watershed[Ahmadi-Nedushan et al., 2006, Saraeva and Hardy, 2009a, Schweizer et al., 2007].52According to Conallin et al. [2010] for an aquatic habitat model to be consideredby water practitioners for current use the following criteria must be met:1. clearly demonstrates the links between the physical data and biological re-quirements using easily obtainable data;2. inputs and results are transparent (i.e. can be presented to different stake-holders)3. the model is user-friendly (i.e. shouldn?t require intensive training)4. have a large spatial scale and can be applied for many different aquaticspeciesPHABSIM and 2-dimensional models require intensive site specific data collectionand technical expertise which often leads to these models being dismissed by prac-titioners. More simplistic and transparent aquatic habitat models are needed.Presented below is a user-friendly, low-input aquatic habitat model that gener-ates WUA as a function of flow. The model uses a low input regime model to setthe reach average bankfull channel conditions from which a reach average cross-sectional shape is inferred. The water level (flow) is sequentially lowered frombankfull conditions with hydraulic properties being recalculated for every itera-tion. The hydraulic properties are then used with a joint frequency depth-velocitydistribution which are combined with applicable HSI to generate WUA for a rangeof species and life stages. Furthermore, future channel conditions and thus habitatcan be predicted for different climate and land use change scenarios by conduct-ing sensitivity analyses. The model outputs are not intended to be treated as the?truth?. However, the model can be used to conduct preliminary assessments ofchannel altering projects and the outputs can help assess future research needs aswell as determine if in-depth habitat assessments (e.g.PHABSIM) are justified.4.2 Rational regime modelRegime theories use optimality criteria over a suitably long time scale to determineequilibrium channel geometry [Kirkby, 1976, Chang, 1979, White et al., 1982,Huang et al., 2004, Nanson and Huang, 2008, Eaton et al., 2010b]. The physically53Figure 4.1: The assumed channel geometry and characteristic rooting depth,H, embedded within UBCRM.based University of British Columbia Regime Model (UBCRM) was chosen as therational regime model to predict reach average bankfull channel dimensions [Eatonet al., 2004]. It has been developed and continually refined through the collabora-tion of researchers in the Department of Geography and the Civil Engineering De-partment at UBC. UBCRM requires modest data inputs thus making it more usefulto environmental practitioners than data intensive, numerically demanding models.Furthermore, the extremal hypotheses embedded within UBCRM are easily under-stood and have been tested against flume and field data. The inclusion of a simpleyet useful bank strength criteria has led to agreement between observed channelform and model predictions [Eaton et al., 2004, Eaton and Church, 2007, Eatonand Millar, 2004].The model determines the channel geometry that allows for the highest system-scale flow resistance for the given inputs [Eaton et al., 2004] as this is deemed themost stable and thus most probable channel configuration [Huang et al., 2004].UBCRM assumes a channel geometry of a cohesionless gravel toe below a verticalupper bank section controlled by a representative rooting depth, H (Figure 4.1).The model requires five user-specified input measures: bankfull discharge (Qb),reach average bed gradient (S), D50, D84, H. These input measures are known forHarris and Fishtrap Creek (refer to Chapter 2). Model outputs were calibrated fromreach average bankfull width (Wb) and depth (db) observed from cross-sectionaldata in the field. If significant deviation occurred between the model outputs andcross-sectional data, H was adjusted until there was agreement between the outputsand observed data.UBCRM can provide current channel geometry as well as approximate channel54dimensions resulting from changes in bankfull flow (Qb) and/or riparian vegetation(H). Thus, future bankfull channel dimensions stemming from out-of-stream wateruse and climate and land use change can be predicted. For the purposes of theproposed aquatic habitat model, UBCRM outputs of interest are the predicted Wband db as they set the reach average bankfull channel dimensions and are essentialfor calibration purposes.As with all models there are some limitations to UBCRM. First, UBCRM formu-lates channel dimensions for an idealized system which the input parameters cansufficiently describe the system. Natural fluvial systems are complex and the influ-ence of hillside processes, large wood (LW), channel sinuosity, heavily armouredbed, etc. can cause large deviations in predicted and observed values. Second, themodel assumes the system is in equilibrium. However, many natural systems arenot stable and exhibit ongoing reach average channel change. For more detaileddescription of the limitations of UBCRM refer to Eaton et al. [2004] and Eaton[2006].4.3 At-a-station hydraulic geometry simulatorThe at-a-station hydraulic geometry simulator (ASHGS) was developed followingthe approach proposed by Ferguson [2003] to incorporate lateral variation in bed-load transport in 1-dimensional models of longitudinal profile development. Inmost natural channels flow strength varies across the channel due to changes indepth, presence of bank friction, or from flow impediments upstream. The Fergu-son [2003] model looked at the variation of shear stress and thus the variation indepth in a 1-dimensional model as means of calculating total bed load flux. Fergu-son [2003] found that simply applying the average depth across the whole channelleads to an underestimation of bedload flux. For the purposes of the proposedhabitat model we are not interested in the variation of shear stress but that of depth.In channels that have approximately a rectangular geometry the lateral varia-tion in depth is negligible. However, for channels that exhibit pool-riffle morphol-ogy this cannot be assumed [Ferguson, 2003]. The variance around the averagedepth is a defining characteristic of these morphologies and has huge implicationsfor aquatic habitat [Eaton et al., 2006, Saraeva and Hardy, 2009a, Schweizer et al.,55Figure 4.2: Cumulative distribution functions of depth for b values of 0.2,0.5, and 0.8. Note the dashed line represents average depth at banfullflow, db2007].All modelling was completed using Matlab (version 7.9.0) statistical programme.To begin, Wb and db are determined using UBCRM. The cumulative depth probabil-ity distribution is determined from an index of channel shape (b). Depths are lessthan db in proportion b of Wb. In this section, b, depths increase linearly from 0to db. In the remaining portion of the channel, 1 - b, the depths are greater than dband increase linearly from db to db/(1 - b) [Ferguson, 2003]. The cumulative depthdistributions for different channel geometries (b-values) are found in Figure 4.2.Each cumulative distribution contains two linear segments that make a dog-leg shape, except when b = 0.5 which contains a single uniform linear segment.The total cross section area (Wb x db) remains the same for all b values. Smallb values are representative of canal-like conditions which exhibit small variationin depth and the maximum depth (dmax) is slightly greater than db. High b valuesrepresent increasingly non-uniform channel conditions with a large proportion ofdepths below db (e.g. channel margins and bars) and dmax is much greater than db.56Often these channel shapes are associated with large meandering channels, braidedchannels, and the presence of multiple thalwegs [Ferguson, 2003]. The resultingcumulative distributions for b ? 0.7 seem highly unlikely in natural channels andb-values of 0.1 to 0.5 seem more representative of fish-bearing channels in BC.Ferguson estimated b by matching moments with fitted distributions of shearstress and water depth. A simpler way of estimating b is desirable for the proposedmodel. Estimating b linearly from Wb/db is proposed:b =Wb/db100(4.1)A linear relationship between b and Wb/db is a major assumption of the modelbut some conceptual and physical evidence is provided. Plane bed channels willhave small Wb/db (close to 10) due to their narrow rectangular form. These channelshave minimal lateral variation in flow and thus depth [Montgomery and Buffington,1997], resulting in b values approaching 0. Wider gravel bed rivers contain bars andexhibit pool riffle morphology. These channels have significant lateral variation inflow [Church, 2006] resulting in a skewed cumulative depth distribution. Verylarge single thread meandering channels can reach Wb/db of close to 60 [Eatonet al., 2010b]. The resulting cumulative depth distribution in these large channelsis estimated at 0.6. Thus, increased width leads to a larger proportion of bars,pools, and riffles resulting in more lateral variation in depth.Furthermore, Ferguson [2003] determined b ranged from approximately 0.4 to0.7 for the lower unconfined Fraser River, British Columbia. Published Wb/db forthis section of Fraser River range from 43 for stable sand beds to 78 for unstablebeds [Desloges and Church, 1989, Rice et al., 2009], thus b values of 0.43 to 0.78using Equation 4.1. Channels with Wb/db greater than 60 (estimated b = 0.6) arelikely to form multiple threads [Eaton et al., 2010b]. The remainder of this chapterwill deal with b values less than 0.5 as this simplifies the model and is representa-tive of the majority of fish bearing channels in British Columbia.Once the cumulative depth distribution and thus reach average cross-sectionalgeometry is established for the reach, the distribution is divided laterally into incre-ments of 0.0001 x Wb. A bankfull water elevation is imposed onto the cross-section(i.e. a vertical line across the top of the distribution). The depth (d) at each of the57lateral increments is determined by subtracting the elevation of the bed from thecumulative depth distribution. A wetted perimeter (Pwetted) is calculated as the lin-ear length of the bed determined to be under water. The hydraulic radius (R) isdetermined using the following equation:R =AxsectionPwetted=W ? d?Pwetted(4.2)where W is the wetted width and d? is the average depth. At bankfull flow the wettedarea is simply Wb x db. The reach average resistance (Res) of the channel for theimposed water level is determined from the following resistance law:Res =a1 ?a2 ? (R/D84)?a21 +a22 ? (R/D84)5/3(4.3)where a1 = 6.5, a2 = 2.5 [Ferguson, 2007]. This particular resistance law was usedas it has the lowest prediction error of the well known resistance laws and it wasdeveloped by the same researcher that proposed the cumulative distributions ofdepth [Ferguson, 2003] which allows for consistency in the proposed model. Theaverage velocity (v?) in the cross section was calculated as follows:v? = Res?g ?R ?S (4.4)where g = 9.81 m s?2. The discharge (Q) associated with the imposed water levelis calculated according to the law of continuity:Q = v? ?Axsection (4.5)Q will equal Qb (or at least very close to) for the first iteration (bankfull conditions).Finally, the Froude number (Fr) is calculated from the following equation:Fr =v??g ? d?(4.6)Once Awetted , Pwetted , R, Res, v?, Q, and Fr are calculated for bankfull water levelthe water level is sequentially dropped by 0.001*dmax and these measures are re-calculated for the new values of W and d?. This process repeats itself until the water58elevation is just above the elevation of dmax (i.e. very low flow).4.4 Application of a habitat modelA joint frequency depth-velocity distribution [Schweizer et al., 2007] was chosenas the statistical habitat method to be used alongside UBCRM and ASHGS. Formore information on statistical habitat methods and rationale for choosing the jointfrequency distribution refer to Chapter 3. The joint frequency distribution utilizes asingle mixing parameter (Smix) to predict both the relative velocity (v/v?) and depthdistributions (d/d?) using a mixture of normal (N) and log-normal (LN) probabilitydensity functions according to the following equations:f (v/v?,Smix) = (1?Smix) ?Nv(?vN ,?vN)+(Smix) ?LNv(?vLN ,?vLN) (4.7)where ?vN = ?vLN = 1, ?vN = 0.52, ?vLN = 1.19, andf (d/d?,Smix) = (1?Smix) ?Nd(?dN ,?dN)+(Smix) ?LNd(?dLN ,?dLN) (4.8)where ?dN = ?dLN = 1, ?dN = 0.52, ?dLN = 1.09, andln(Smix1?Smix)=?4.72?2.84 ? ln(Fr) (4.9)All inputs needed for the joint-frequency distribution are determined using ASHGS.The distribution equations were examined for 0 < v/v? < 3 and 0 < d/d? < 3 using30 equidistant bins.The absolute values of the bins were determined by multiplying the bins of therelative hydraulic distributions by their respective mean. Subsequently, the abso-lute bin values were compared to HSI to determine the suitability of each bin. Thetotal suitability of the empirical velocity and depth distributions was determinedfrom the following equations:59Velocity Suitability =n?i=1bin frequency ?bin suitabilityn?i=1bin frequency(4.10)Depth Suitability =n?i=1bin frequency ?bin suitabilityn?i=1bin frequency(4.11)where n is the total number of bins in each distribution. Upon calculating thesemeasures the WUA was calculated from the following equation:WUA = W ?Velocity Suitability ?Depth Suitability (4.12)For this chapter WUA will be evaluated as an area per unit length (m2 m?1) toavoid having to estimate the wetted area. The habitat value (HV), a dimensionlessmeasure that allows for easy comparison between reaches, was calculated from thefollowing equation:HV = Velocity Suitability ?Depth Suitability (4.13)Habitat indices calculated using the statistical distributions do not contain any spa-tial explicit detail. Hydraulic distributions are determined for every water levelobserved with ASHGS. Thus, WUA(Q) and HV (Q) are determined across a rangeof possible flow conditions.4.5 Model testingThe following are the calibrated input values used for Harris Creek: Qb = 19 m3s?1, S = 0.011 m m?1, D50 = 68 mm, D84 = 135 mm, H = 0.35 m. UBCRM predictedreach average Wb = 14.36 m, db = 0.705 m and v? = 1.88 m s?1. These values arequite similar to an observed mean Wb = 14.57 m (mean of 12 cross sections) anddb = 0.702 m and v? = 1.866 m s?1 produced from a River2D simulation. UsingEquation 4.1, a b-value of 0.203 was estimated for Harris Creek (Figure 4.3). Thisvalue seems reasonable as Harris Creek has a relatively plane bed upper reach (b60Figure 4.3: Predicted reach average channel geometry at Harris Creek (b =0.203).values closer to 0.1) and a pool-riffle morphology in the mid to lower reach (brange from 0.15 to 0.5).The v? and d? values produced by ASHGS were compared to the v? and d? producedin 18 River2D simulations (Tables 4.1 and 4.2). d? had a maximum deviation of -0.011 m (6.08 %) at Q = 0.75 m3 s?1 and a mean deviation of 1.06 %. In general,ASHGS over-predicted d? at low flows and under-predicted d? at high flows (Table4.1). v? had a maximum deviation of -0.077 m s?1 (16.96 %) at Q = 0.75 m3 s?1and a mean deviation of 4.47 %. ASHGS under-predicted v? at low flows and over-predicted at high flows (Table 4.2). The larger deviation associated with v? is likelydue to the different flow resistance laws utilized by River2D and ASHGS. Therealways exists uncertainty in the input parameters which manifests as predictionerror in flow resistance equations. Also, large relative errors are inherent duringconditions of partial submergence (R/D84 <1), which occurred during low flowconditions at Harris Creek [Ferguson, 2007]. Furthermore, UBCRM was calibratedto bankfull conditions. As flow simulations move further from bankfull conditionsthe prediction error is likely to become higher.61Table 4.1: Modelled d? using River2D and ASHGS for a range of flows atHarris CreekQ (m3 s?1) d?River2D (m) d?ASHGS (m) Difference0.75 0.181 0.170 -0.011 (6.08 %)1.00 0.200 0.192 -0.008 (4.00 %)1.23 0.214 0.209 -0.005 (2.34 %)1.32 0.219 0.216 -0.003 (1.37 %)1.61 0.236 0.234 -0.002 (0.85 %)1.93 0.254 0.254 0 (0 %)2.29 0.272 0.273 0.001 (0.37 %)2.61 0.290 0.289 -0.001 (0.34 %)3.08 0.308 0.310 0.002 (0.65 %)3.51 0.328 0.328 0 (0 %)3.98 0.346 0.347 0.001 (0.29 %)4.47 0.365 0.366 0.001 (0.27 %)5.00 0.385 0.384 -0.001 (0.26 %)5.55 0.401 0.401 0 (0 %)7.50 0.458 0.460 0.002 (0.44 %)10.00 0.519 0.525 0.006 (1.16 %)15.00 0.632 0.632 0 (0 %)19.00 0.707 0.702 -0.005 (0.71 %)The HV generated by the proposed aquatic habitat model was compared to HVproduced from River2D and from a statistical habitat model (developed in Chapter3) using a joint frequency distribution developed by Schweizer et al. [2007]. TheSchweizer et al. [2007] habitat model utilized reach average hydraulic conditionsproduced in the River2D simulations. HV were chosen as the means of comparisonbecause it allows for a direct unbiased comparison (WUA relies on wetted widthor wetted area which could be different between River2D and the proposed habitatmodel).Three species were used for the comparison: Oncorhynchus mykiss (rainbowtrout), Micropterus dolomieu (smallmouth bass), and Rhinichthys cataractae (long-nose dace). Adult, juvenile, and spawning rainbow trout life stages were examined.Depth and velocity preferences were obtained from United States Geological Sur-62Table 4.2: Modelled u? using River2D and ASHGS for a range of flows atHarris CreekQ (m3 s?1) v?River2D (m s?1) v?ASHGS (m s?1 ) Difference0.75 0.454 0.377 -0.077 (16.96 %)1.00 0.508 0.441 -0.067 (13.19 %)1.23 0.552 0.494 -0.058 (10.51 %)1.32 0.565 0.516 -0.049 (8.67 %)1.61 0.616 0.571 -0.045 (7.31 %)1.93 0.667 0.630 -0.037 (5.55 %)2.29 0.718 0.690 -0.028 (3.90 %)2.61 0.767 0.739 -0.028 (3.65 %)3.08 0.815 0.804 -0.011 (1.3 %)3.51 0.867 0.859 -0.008 (0.92 %)3.98 0.917 0.915 -0.002 (0.22 %)4.47 0.966 0.970 0.004 (0.41 %)5.00 1.019 1.025 0.006 (0.59 %)5.55 1.066 1.072 0.006 (0.56 %)7.50 1.219 1.243 0.024 (1.97 %)10.00 1.381 1.342 0.039 (2.82 %)15.00 1.671 1.696 0.025 (1.50 %)19.00 1.858 1.866 0.008 (0.43 %)vey (USGS) HSI. For more information on each species? hydraulic preferencesrefer to Chapter 3 and Appendix A. Comparison of HV for the species and lifestages under consideration are found in Figures 4.4 though 4.8.In general, ASHGS over-predicted HV in comparison to River2D. This is mostlikely due to the different ways each model calculate HV (refer to Chapter 3). Formost species and life stages HV predicted by ASHGS were very similar to that ofthe Schweizer et al. [2007] model. Both use the same empirical hydraulic distribu-tions but the Schweizer et al. [2007] model uses hydraulic conditions generated byRiver2D whereas ASHGS predicts the hydraulic conditions. Similar predicted HVfrom the two models further suggests ASHGS is adequately predicting hydraulicconditions across a range of flows.63Figure 4.4: HV produced by River2D, Schweizer et al., and ASHGS for adultrainbow trout.Figure 4.5: HV produced by River2D, Schweizer et al., and ASHGS for juve-nile rainbow trout.64Figure 4.6: HV produced by River2D, Schweizer et al., and ASHGS forspawning rainbow trout.Figure 4.7: HV produced by River2D, Schweizer et al., and ASHGS for adultsmallmouth bass.65Figure 4.8: HV produced by River2D, Schweizer et al., and ASHGS for adultlongnose dace.4.6 Sensitivity analyses4.6.1 Harris CreekOften the goal of aquatic habitat modelling is prediction. Of particular inter-est is the influence of climate and land use change on aquatic habitat in BritishColumbian channels. British Columbia is predicted to have greater warming andchanges in precipitation patterns than the global average [MFLNR, 2009]. Warm-ing will occur in all seasons but will be greatest in the winter. Winters are expectedto be wetter across British Columbia and summers will be drier in Southern andCentral British Columbia and wetter in Northern British Columbia.Increases in the frost free period, reduced snow packs, and earlier spring melt-ing will alter the hydro-graph and influence Qb. Increases in evaporative demandof the atmosphere and frequency of extremely warm days could potentially leadto a reduction in the mean minimal monthly, weekly, and daily flow which couldcompromise the biological integrity of the channel [Dakova et al., 2000, Stalnaker,66Table 4.3: Predicted changes in climatic variables at Harris Creek under theCanadian Center for Climate Modelling and Analysis A2 ScenarioVariable 1971 - 2000 2050 2080Mean Annual Temperature ( ?C) 5.8 8.2 9.7Mean Winter Temperature ( ?C) -4.6 -2.1 -0.8Mean Summer Temperature ( ?C) 16 18.4 19.9Maximum Mean Summer Temperature ( ?C) 23.8 26.5 28.2Mean Annual Precipitation (mm) 546 581 577Mean Summer Precipitation (mm) 250 246 216Precipitation as snow (mm) 144 98 631979, Tennant, 1976].There exists many future green house gas emissions scenarios. Two of themost well known are the A2 and B1 emissions scenarios. Under the A2 scenario,there will be little success at curbing future global emissions leading to a 3 to5 ?C warming in British Columbia by 2080. The B1 scenario is representativeof a substantial reduction in global emissions leading to a 2 to 3 ?C warming inBritish Columbia by 2080 [MFLNR, 2009]. Some climatic variables that are ofimportance to flow regime and channel conditions are summarized in Table 4.3 (A2scenario) and Table 4.4 (B1 scenario). These measures are predicted for a locationjust upstream of the study site at Harris Creek using ClimateBC Map which wasdeveloped by The Centre for Forest Conservation Genetics at the University ofBritish Columbia. The measures are based on General Climate Models producedat the Canadian Center for Climate Modelling and Analysis.Warming temperatures and a changing precipitation regime will influence ri-parian vegetation dynamics. Furthermore, forest fire intensity and frequency willincrease due to higher fuel loads under a warming and drier climate [MFLNR,2009]. Forest fires influence channel morphology by significantly reducing bankstrength, increasing the amount of in-channel LW, altering the timing and magni-tude of peak flows, and modifying the volume and character of sediment deliveredto the channel [Eaton et al., 2010a]. As well, demand for riparian areas as futureagricultural, industrial, and recreational sites will lead to changes in riparian vege-67Table 4.4: Predicted changes in climatic variables at Harris Creek under theCanadian Center for Climate Modelling and Analysis B1 ScenarioVariable 1971 - 2000 2050 2080Mean Annual Temperature ( ?C) 5.8 7.6 8Mean Winter Temperature ( ?C) -4.6 -2.5 -2.2Mean Summer Temperature ( ?C) 16 18 18.5Maximum Mean Summer Temperature ( ?C) 23.8 26 26.4Mean Annual Precipitation (mm) 546 577 612Mean Summer Precipitation (mm) 250 256 266Precipitation as snow (mm) 144 105 103tation along British Columbian channels. How land use change and climate changewill interact and thus influence riparian vegetation in British Columbia is difficultto predict.To investigate the potential influence of climate and land use change on channeldimensions and aquatic habitat, Qb and H were varied one at a time to predict futurechannel trajectories for Harris Creek. Changes in S as well as sediment supply andtransport are to some extent stochastic processes and will be excluded from thesensitivity analysis. Qb was examined over a range of ? 50 % of its current value(Qbo). Changes in Qb due to climatic factors or out-of-stream uses are likely to fallwithin this range. H was examined from - 80% to + 20% of its current value (Ho).Bank strength can fall as low as - 80% of its pre-fire value in the decade followingthe fire as seen by a model presented by Benda and Dunne [1997]. H was examinedup to + 20% of Ho to allow for potential maturation of the rooting system at HarrisCreek, which could be a possibility under a warmer, wetter climate.The model was calibrated to Harris Creek with the same input values as pre-sented in the previous section. The sensitivity of channel dimensions to variableQb as determined using UBCRM can be seen in Figure 4.9. The width changes by amaximum of about ? 40 %, while db stays relatively constant across the range ofQb. Subsequently, the Wb/db and thus b follows the same trajectory as Wb. Both theA2 and B1 scenarios indicate a reduced snow pack in the Harris Creek watershed,which will most likely lead to a decrease in Qb. Under these scenarios the chan-68Figure 4.9: Response of Wb, db, and b to changes in Qb at Harris Creek.nel would become narrower, depth would stay relatively constant, and the channelwould have a more canal like structure and less pool-riffle units (smaller b value).If Qb were to increase due to increased severe storm activity or deforestation in thewatershed it is predicted that the channel would become wider, the depth wouldremain relatively constant, and the channel would establish larger bars and a moreprominent pool-riffle morphology (larger b value).Moreover, the sensitivity of channel dimensions to variable H as determinedusing UBCRM can be seen in Figure 4.10. Under scenarios where the riparian veg-etation becomes slightly more mature, the channel will become narrower, deeper,and have a more canal-like structure. More likely scenarios are a decrease in rootstrength due to land use change, most notably from a conversion to agricultural landin the Harris Creek watershed, or forest fire. In the years following a forest fire Wbwould increase by approximately 70% and db would decrease by about 25%. Thisleads to b increasing by up to 120%, which indicates a wandering channel withmany bars and potentially multiple thalwegs.The sensitivity of WUA and HV to changes in Qb were also examined for the69Figure 4.10: Response of Wb, db, and b to changes in H at Harris Creek.species and life stages mentioned above (See Figures 4.11 to 4.13 and AppendixC). Note that mean annual flow was set at 1.5 m3 s?1 and minimum mean monthlyflow was estimated to be 0.5 m3 s?1. In general, WUA curves moved upward asQb increased. This is because there is more wetted area with increased Qb due toa wider channel geometry which leads to more habitat becoming available. Therewas little difference in adult smallmouth bass WUA across the range of Qb ex-amined because Harris Creek is very poor habitat for this particular species. HVremoves the bias of wetted area. As Qb increases the HV curves move to the right.This is because as Qb increases the channel becomes wider. Thus, higher flowsare needed to raise the water level to habitable levels. A reduction in Qb, which islikely under a warmer climate results in decreased habitat availability for rainbowtrout and longnose dace.Furthermore, the sensitivity of WUA and HV to changes in H produced moredrastic morphologic change (see Figures 4.14 to 4.16 and Appendix C). In general,decreasing H led to the WUA curves shifting upwards. This is due to the increasein wetted area from the widening channel. 0.2Ho does not follow this general trend70Figure 4.11: Adult rainbow trout WUA and HV at Harris Creek across arange of potential Qb. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure 4.12: Adult smallmouth bass WUA and HV at Harris Creek across arange of potential Qb. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.71Figure 4.13: Adult longnose dace WUA and HV at Harris Creek across arange of potential Qb. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.for rainbow trout because the channel becomes too wide and shallow. HV curvesshift to the right as H decreases as higher flows are needed to bring the water levelto habitable levels in the widened channel. Thus, a reduction in bank strength atHarris Creek due to forest fire or deforestation can increase the amount of availablehabitat for longnose dace and rainbow trout provided it doesn?t cross a threshold.The sensitivity of WUA at mean annual flow and minimum mean monthly lowflow were also examined as these flows are of biological importance (see Figures4.17 to 4.19 and Appendix C). The sensitivity of these measures are species spe-cific. As well, the maximum WUA was also examined across a range of Qb andH. As expected, maximum WUA for the most part increases with increasing Qbvalues and increases with decreasing H values as the channel becomes wider andmore wetted area becomes available. However, at very small H values the channelbecomes too wide for the flow regime to produce habitat water levels for rainbowtrout, which leads to a drop in maximum WUA at low H values.72Figure 4.14: Adult rainbow trout WUA and HV at Harris Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure 4.15: Adult smallmouth bass WUA and HV at Harris Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.73Figure 4.16: Adult longnose dace WUA and HV at Harris Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure 4.17: Sensitivity of adult rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of Qb and H.74Figure 4.18: Sensitivity of adult smallmouth bass WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maximumWUA for a range of Qb and H.Figure 4.19: Sensitivity of adult longnose dace WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of Qb and H.754.6.2 Fishtrap CreekThe high-intensity forest fire that burned most of the riparian vegetation at FishtrapCreek in 2003 provides interesting sensitivity analysis. The model was calibratedto Fishtrap Creek using the following input values: Qb = 7.5 m3 s?1, S = 0.02m m?1, D50 = 55 mm, D84 = 128 mm, H = 0.46 m. The mean annual flow wasset at 1 m3 s?1 and minimum mean monthly flow was estimated to be 0.5 m3s?1. The predicted channel width is 9.4 m, which is comparable to the 9.5 mmeasured pre-fire width. The pre-fire predicted b value is 0.162, indicating smallbut significant lateral variation in depth and velocity. Qb has remained close tohistorical values following the fire [Eaton et al., 2010a] and it will be held constantduring the analyses.It is predicted from a model presented by Benda and Dunne [1997] that Hwill fall to as low as 20% of its pre-fire value in the decade following the fire.This drastic drop in H will lead to significant widening of the channel (Figure4.20). Fishtrap Creek will become shallower due to the widening and aggradationthat will occur. As well, b is predicted to increase by up to 300% of its originalvalue, indicating huge lateral variation resulting from the development of bars andmultiple channel threads [Eaton et al., 2010b]. Channel widening, aggradation,and the development of bars and multiple thalwegs were observed (although not tothe extreme that the sensitivity analysis suggests) at Fishtrap Creek beginning from2006 through 2008 [Eaton et al., 2010a,c]. Morphological change is most likelystill ongoing. Deviation between the suggested dimensions from the sensitivityanalysis and observed conditions could be due to the presence and continued influxof LW into the channel as well as changes in the type and volume of sedimentsupplied to the channel. The channel is likely to drift back to pre-fire conditionsand thus channel dimensions and habitat indices will evolve from the extremesassociated with 0.2Ho to the more stable channel configuration associated with Ho.Changes in habitat indices (Figure 4.21 to 4.24 and Appendix C) follow similartrajectories as habitat indices at Harris Creek. In general, a decrease in H resultsin increased WUA for adult, juvenile, and spawning rainbow trout as the chan-nel became wider and shallower. The exception being 0.2Ho conditions when thechannel becomes very wide and shallow for adult and juvenile life stages resulting76Figure 4.20: Response of Wb, db, and b to changes in H at Fishtrap Creek.in anomalously poor habitat measures. This suggests the post-fire channel couldpresent habitat conditions that are too extreme for adult and juvenile rainbow trout.Habitat remains poor for smallmouth bass over the range of H values examinedand thus their habitat indices did not change significantly. Longnose dace WUAincreased with decreasing H suggesting Fishtrap Creek is the most habitable forlongnose dace when there is a significant reduction in bank strength.4.7 Model limitationsThe proposed habitat model will only be as good as the ability of each of thethree components (UBCRM, ASHGS, statistical habitat method). Erroneous predic-tions by UBCRM will result in inaccurate predictions by ASHGS and the statisticalhabitat methods. Limitations of UBCRM are briefly highlighted in Section 4.2 anddiscussed in detail in Eaton et al. [2004] and Eaton [2006]. Similarly, if ASHGSis unable to model observed mean hydraulic conditions there will be errors in thehabitat indices. Moreover, ASHGS must be validated on more channels of varyingsize and ecological regimes. The outputted reach average hydraulic conditions andhabitat indices have only been tested at Harris Creek.77Figure 4.21: Adult rainbow trout WUA and HV at Fishtrap Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure 4.22: Sensitivity of adult rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of H at Fishtrap Creek78Figure 4.23: Adult longnose dace WUA and HV at Fishtrap Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure 4.24: Sensitivity of adult longnose dace WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of H at Fishtrap Creek.79Furthermore, assuming b and Wb/db are linearly related is a major and untestedassumption of ASHGS and is an avenue that requires further research before ASHGScan be used as a legitimate aquatic habitat tool. It is most likely that b is dependenton many factors in a fluvial system that were overlooked by this model. Greaterunderstanding of how b changes longitudinally along a reach, between differentmorphological units, and between channels is not only of benefit to the proposedmodel but to fluvial ecological modelling as a whole. As well, greater understand-ing on how to estimate b in multi-thread channels is needed, as the proposed modelonly deals with single thread channels.The proposed model provides predictions of reach average hydraulic condi-tions and lateral depth probability distributions. However, it does not provide anyinformation on the spatial distribution of hydraulic variables (unlike 2-dimensionalmodels) and thus does not distinguish between different habitat units. Lumping allthe unique habitat units of a channel into a reach average geometry can be a hugesimplification of complex channel structures and could lead to erroneous habitatquantification and overlook potential habitat. For instance, a channel that has bothplane bed and pool-riffle morphology, such as Harris Creek, will be inferred as achannel that has intermediate morphology between those two end members. Thus,the presence of an important habitat unit such as a deep pool will be reduced dueto the simplification present within the model. As well, the model ignores thepresence of LW and lateral variation in sediment type, both of which are hugelyimportant to aquatic habitat [Ahmadi-Nedushan et al., 2006].Finally, the statistical habitat method embedded in the proposed model worksreasonably well at Harris Creek. However, the transferability of the statisticalmethods to other channels should be done with caution as they were developed onNew Zealand channels that exhibit a relatively natural flow regime and are meantfor flow conditions that are below mean annual discharge [Schweizer et al., 2007].As well, the statistical habitat methods would benefit from examining a wider rangeof species and life stages. Examining how the model fairs between young and oldlife stages and between pelagic and benethic species would be very useful. For thisto be done a larger database of reliable HSI for British Columbian channels needsto be established and become easily accessible to practitioners. Furthermore, per-haps ASHGS would benefit from the use of non-statistical and/or non-HSI based80habitat models such as process-oriented bioenergetic models or fuzzy-rule basedmodels [Anderson et al., 2006]. For more detailed information on limitations ofthe statistical habitat methods and habitat quantification in general refer to Chapter3.4.8 ConclusionThe proposed aquatic habitat model combines a low-input regime model (predictsreach average bankfull channel dimensions) with ASHGS to model reach averagehydraulic variables across a range of relevant flows. The reach average hydraulicvariables are inputted into statistical habitat models to produce hydraulic distribu-tions which are applied to HSI to generate habitat indices. Hence, the proposedmodel allows users to predict aquatic habitat indices for different fish species andlife stages without the need for extensive data collection and painstaking analysis.Due to its low-input nature, transparency, user-friendliness, and large spatial appli-cability, the proposed model has the potential to become an accepted tool amongstpractitioners [Conallin et al., 2010].The use of a regime model allows for input variables to be easily varied. Thus,the proposed model provides the framework to quickly and adequately examineaquatic habitat and hypothesize how a changing channel structure will influencehydraulic habitat. To the author?s knowledge it is the first aquatic assessment toolto directly link climate and land use change to aquatic habitat. As well, ASHGS canbe used to evaluate longitudinal changes in aquatic habitat by simply knowing howQ, S, characteristic grain size, and rooting depth change downstream.A reduction in Qb, which is likely to occur at Harris Creek and Fishtrap Creekunder a warming climate, results in narrower and deeper channels. This will lead toa reduction in WUA for all species under investigation which suggests practitionershave to be mindful of future flow abstractions at these channels as climatic factorsare likely to decrease available habitat over the coming century. As well, a decreasein bank strength due to deforestation or forest fire results in a wider, shallowerchannel which leads to increases in WUA for longnose dace and rainbow trout.However, at very low bank strengths the channel becomes too wide and shallowfor some life stages of rainbow trout resulting in decreased habitat availability.81Thus, loss of bank strength can be beneficial for some species provided it doesn?treach a threshold. Smallmouth bass habitat did not change significantly duringthe sensitivity analyses because hydraulic conditions were always poor for thisparticular species.The proposed model provides the framework of a preliminary assessment orrapid evaluation tool. Due to its low-input nature, the model provides a means ofhabitat assessment where site-specific data does not exist. The model can be usedto evaluate whether anticipated or proposed changes to the channel or riparian areaare approaching a habitat threshold and thus indicate the need for more exten-sive habitat assessment (e.g. PHABSIM, 2-dimensional hydrodynamic models). Aswell, it is useful when assessments need to be conducted on multiple reaches andchannels (e.g. basin-wide habitat assessment). Further research is needed on pre-dicting reach average channel geometry (b) as well as determining the influence ofLW and variable sediment supply on habitat indices before this model can becomea legitimate aquatic habitat tool.82Chapter 5Conclusions5.1 Empirical hydraulic distributions in BritishColumbian channelsEmpirical hydraulic distribution equations were evaluated on two channels in theInterior Region of British Columbia. Measured hydraulic distributions were ad-equately reproduced for both high and low flow conditions at Harris Creek us-ing statistical distributions. Likewise, statistical velocity distributions were able torecreate the measured velocity distribution for both 2006 and 2007 high flow dataat Fishtrap Creek, a channel recently disturbed by forest fire. Empirical depth dis-tributions were unable to model the measured depth distributions at Fishtrap Creekfollowing rapid morphological change. The empirical distributions were developedon channels that had relatively undisturbed morphologies suggesting their use onrecently disturbed channels should be done with caution [Schweizer et al., 2007].Furthermore, the empirical hydraulic equations were compared to depth andvelocity distributions produced by a 2-dimensional hydrodynamic model (River2D).Depth and velocity distributions were sufficiently recreated by statistical distribu-tions at flows < 3 m3 s?1. The empirical equations provided strong fits to theRiver2D hydraulic distributions at flows close to and below mean annual flow.As flows approached bankfull (19 m3 s?1), empirical distributions were unable tomodel the high densities surrounding v? and d?. In particular, the empirical velocityequations performed very poorly at high flows. These findings concur with re-83sults of a similar investigation in the Nooksack River basin [Saraeva and Hardy,2009a]. Thus, empirical equations are most appropriate at low flow conditions,which are the limiting flows for many aquatic species [Dakova et al., 2000, Hat-field and Bruce, 2000].A joint frequency depth-velocity empirical distribution [Schweizer et al., 2007]was deemed most suitable for modelling both measured and simulated hydraulicdistributions. The joint frequency distribution was paired with HSI to create a low-input aquatic habitat model that generates habitat indices across a range of flowsfor three species at Harris Creek. WUA produced by the low-input statistical habitatmodel and River2D (data-intensive 2-dimensional model) compared favourably atlow flows. There existed deviation in the absolute values of WUA at high flows.However, the general shape and trends in WUA data simulated by River2D werereproduced by the statistical habitat model.Empirical hydraulic equations can be useful for future aquatic habitat mod-elling endeavours in British Columbian channels. In particular, the results showthat depth and velocity distributions are adequately recreated at low flow condi-tions. With knowledge of the channel and expertise with in-stream flow method-ologies, empirical hydraulic distributions alongside HSI can be useful for prelim-inary assessments and basin wide habitat studies, especially when environmentaldata is lacking.As with all in-stream flow methodologies there are limitations that need tobe acknowledged or addressed. The empirical distributions do not provide anyinformation on the spatial distribution of the predicted hydraulic variables. As well,the proposed habitat model does not incorporate substrate, cover, and wood loaddata, all of which determine aquatic habitat [Allan and Castillo, 2007]. Finally,for the statistical habitat methodology to be improved there needs to be a hugeimprovement in the predictive capacity of the empirical velocity equation.5.2 ASHGS aquatic habitat modelA channel regime model (UBCRM) was paired with ASHGS and a statistical habi-tat model to predict habitat indices across a range of flows for species and lifestages found in British Columbian channels. The proposed methodology is a sim-84ple alternative to more widely used data-intensive in-stream flow methodologies(i.e. PHABSIM and 2-dimensional hydrodynamic models). Furthermore, futurereach average channel dimensions resulting from climate and land use change canbe modelled with UBCRM which allows for future habitat indices to be predictedusing the ASHGS aquatic habitat model. Thus, the proposed methodology providespractitioners with a simplistic tool that can highlight future channel dimensionsand flow regimes of concern and illustrate ecological thresholds.In British Columbia, reduced snowpacks and earlier spring melting caused by awarming climate is likely to reduce Qb in channels across the province. A decreasein Qb will lead to narrower, deeper channels which will in turn reduce availablehabitable area for many species found in British Columbian channels. Further-more, decreases in bank strength due to deforestation of the riparian vegetation andforest fire will lead to wider and shallower channels in British Columbia. Widerchannels can be beneficial to many fish species as the wetted area will increase.Severe reduction in bank strength (e.g. 5 -10 years following a forest fire) canlead to extremely wide channels causing hydraulic habitat conditions to becometoo extreme for some species.The proposed ASHGS aquatic habitat model is ideal for preliminary assessmentand basin-wide studies. The model should be accepted by practitioners as it istransparent, user-friendly, and can be used on a wide range of channels [Conallinet al., 2010]. The required input data are often previously established for a channelor can be obtained within one day of field work. Future research and refinement ofthe ASHGS habitat model is needed for the methodology to become a legitimate in-stream flow assessment tool. First and foremost, the model needs to be evaluated onchannels of varying flow regimes and morphologies across the province. As well,the model?s performance at predicting habitat indices for different fish species andlife stages is imperative. A stronger understanding of how to predict the reachaverage channel geometry (b value) is crucial as the relationship between channelmorphology and reach average channel geometry is currently poorly understood.855.3 The future of aquatic habitat modelling in BritishColumbiaContinued effort is needed to develop and refine in-stream flow assessment toolsfor British Columbian channels. In particular, greater emphasis needs to be placedon developing low-input, inexpensive tools that are reliable and have a sound sci-entific base [Conallin et al., 2010, Hatfield and Bruce, 2000]. The use of empiricalstatistical hydraulic equations and habitat models that use these equations have tobe further refined in British Columbia. In particular, further evaluation is neededon a wider range of fish-bearing channels.The development of British Columbia specific empirical hydraulic equationscould be quite useful (although it is a rather large undertaking). The proposed em-pirical distributions were developed in different biogeoclimatic zones than thosefound in British Columbia. Manipulation of previous empirical models in theNooksack River basin provided a more reliable in-stream flow assessment tool[Saraeva and Hardy, 2009a]. Furthermore, the inability of the empirical distri-butions to adequately model hydraulic conditions in a small pluvial stream in theCoast Mountains [Rosenfeld et al., 2011] and the depth distribution at a recentlydisturbed channel (Fishtrap Creek) highlights the need for some form of local cali-bration of existing empirical equations or distribution equations that are developedon the unique riverscapes found across British Columbia.Future in-stream flow methodologies need to be able to quantify habitat loss orgain resulting from changes in flow regime, sediment supply, and riparian vegeta-tion [Conallin et al., 2010]. The influence of climate change on these environmentalconditions is poorly known. Having a better understanding of how environmentalconditions will evolve over the coming decades will foster more accurate predic-tions of future flow regimes and channel morphologies and thus future hydraulichabitat availability [Tharme, 2003].Furthermore, low-input flow methodologies tools should examine the influenceof sediment supply and LW on reach average channel conditions. These two vari-ables were ignored in this project as they are difficult to model due to their stochas-tic nature. However, substrate and LW undoubtedly influence channel morphologyas well as the mesohabitats that are present within the channel. Also, incorporat-86ing the distribution of shear stresses can provide more accurate model predictions[Saraeva and Hardy, 2009a]. Within channel shear stress controls substrate distri-bution and type, which in turn influences channel slope and thus the distribution ofdepths and velocities.There has been limited development and use of physical methodologies thatexamine longitudinal changes in habitat along a given channel [Laliberte et al.,2013, Rosenfeld et al., 2007]. The proposed ASHGS habitat model could be usedto examine downstream habitat changes although that is not its intended purpose.Furthermore, aquatic habitat models have long been scrutinized for not incorpo-rating ecological interactions and thus future methodologies would benefit from abroader ecological perspective [Anderson et al., 2006, Conallin et al., 2010].Lastly, it is very possible that the way forward in aquatic habitat modelling inBritish Columbian channels is not empirical distributions or any of the in-streamflow methodologies mentioned in this dissertation for that matter. Other approachessuch as fuzzy-rule based modelling which incorporate ecological interactions andrequire modest data inputs are viable options. Furthermore, there are a host ofother methodologies that have and are continually being developed to provide wa-ter practitioners with the best tools to withhold the ecological integrity of contestedchannels across the globe. Continual evaluation, refinement, and collaboration willprovide the most suitable in-stream flow methodologies moving forward into an eraof uncertain climate.87BibliographyB. Ahmadi-Nedushan, A. St-Hilaire, M. Be?rube?, E?. Robichaud, N. Thie?monge,and B. Bobe?e. A review of statistical methods for the evaluation of aquatichabitat suitability for instream flow assessment. River Research andApplications, 22(5):503?523, 2006. ? pages 1, 7, 8, 52, 80J. D. Allan and M. M. Castillo. Stream ecology: structure and function of runningwaters. Springer, 2007. ? pages 22, 84K. E. Anderson, A. J. Paul, E. McCauley, L. J. Jackson, J. R. Post, and R. M.Nisbet. Instream flow needs in streams and rivers: the importance ofunderstanding ecological dynamics. Frontiers in Ecology and the Environment,4(6):309?318, 2006. ? pages 1, 81, 87T. C. Annear, C. C. Estes, and I. F. Council. Instream flows for riverine resourcestewardship. Instream Flow Council, 2004. ? pages 2, 22C. Armour and J. Taylor. Evaluation of the instream flow incrementalmethodology by US Fish and Wildlife Service field users. Fisheries, 16(5):36?43, 1991. ? pages 5L. Benda and T. Dunne. Stochastic forcing of sediment supply to channelnetworks from landsliding and debris flow. Water Resources Research, 33(12):2849?2863, 1997. ? pages 68, 76K. D. Bovee. Guide to stream habitat analysis using the instream flow incrementalmethodology. Available from the National Technical Information Service,Springfield VA 22161 as PB 83-131052. Report, 1982. ? pages 4, 49, 51K. D. Bovee, B. L. Lamb, J. M. Bartholow, C. B. Stalnaker, and J. Taylor. Streamhabitat analysis using the instream flow incremental methodology. Technicalreport, DTIC Document, 1998. ? pages 688H. H. Chang. Minimum stream power and river channel patterns. Journal ofHydrology, 41(3):303?327, 1979. ? pages 53M. Church. Bed material transport and the morphology of alluvial river channels.Annu. Rev. Earth Planet. Sci., 34:325?354, 2006. ? pages 57M. Church and M. Hassan. Mobility of bed material in Harris Creek. WaterResources Research, 38(11):1237, 2002. ? pages 10, 17J. Conallin, E. Boegh, and J. K. Jensen. Instream physical habitat modellingtypes: an analysis as stream hydromorphological modelling tools for eu waterresource managers. Intl. J. River Basin Management, 8(1):93?107, 2010. ?pages 1, 2, 4, 6, 22, 53, 81, 85, 86, 87D. M. Corbett. Stream-gaging procedure: a manual describing methods andpractices of the geological survey. Geological Survey water-supply paper, 888,1962. ? pages 15S. Dakova, Y. Uzunov, and D. Mandadjiev. Low flow: the river?s ecosystemlimiting factor. Ecological Engineering, 16(1):167?174, 2000. ? pages 40, 47,66, 84S. Day and W. Fletcher. Effects of valley and local channel morphology on thedistribution of gold in stream sediments from harris creek, british columbia,canada. Journal of geochemical exploration, 32(1):1?16, 1989. ? pages 10J. R. Desloges and M. A. Church. Wandering gravel-bed rivers. The CanadianGeographer/Le Ge?ographe canadien, 33(4):360?364, 1989. ? pages 57B. Eaton, M. Church, and R. Millar. Rational regime model of alluvial channelmorphology and response. Earth Surface Processes and Landforms, 29(4):511?529, 2004. ? pages 54, 55, 77B. Eaton, C. Andrews, T. Giles, and J. Phillips. Wildfire, morphologic change andbed material transport at Fishtrap Creek, British Columbia. Geomorphology,118(3):409?424, 2010a. ? pages 18, 20, 33, 67, 76B. Eaton, R. G. Millar, and S. Davidson. Channel patterns: braided,anabranching, and single-thread. Geomorphology, 120(3):353?364, 2010b. ?pages 53, 57, 76B. Eaton, R. Moore, and T. Giles. Forest fire, bank strength and channelinstability: the ?unusual? response of Fishtrap Creek, British Columbia. EarthSurface Processes and Landforms, 35(10):1167?1183, 2010c. ? pages 18, 19,20, 33, 7689B. C. Eaton. Bank stability analysis for regime models of vegetated gravel bedrivers. Earth Surface Processes and Landforms, 31(11):1438?1444, 2006. ?pages 55, 77B. C. Eaton and M. Church. Predicting downstream hydraulic geometry: A test ofrational regime theory. Journal of Geophysical Research: Earth Surface(2003?2012), 112(F3), 2007. ? pages 54B. C. Eaton and R. G. Millar. Optimal alluvial channel width under a bankstability constraint. Geomorphology, 62(1):35?45, 2004. ? pages 54B. C. Eaton, M. Church, and T. R. Davies. A conceptual model for meanderinitiation in bedload-dominated streams. Earth Surface Processes andLandforms, 31(7):875?891, 2006. ? pages 55R. Ferguson. The missing dimension: effects of lateral variation on 1-dcalculations of fluvial bedload transport. Geomorphology, 56(1):1?14, 2003.? pages 55, 56, 57, 58R. Ferguson. Flow resistance equations for gravel- and boulder-bed streams.Water Resources Research, 43(5), 2007. ? pages 58, 61W. Fletcher and J. Wolcott. Transport of magnetite and gold in harris creek,british columbia, and implications for exploration. Journal of GeochemicalExploration, 41(3):253?274, 1991. ? pages 10M. Gard. Comparison of spawning habitat predictions of PHABSIM and River2Dmodels. International Journal of River Basin Management, 7(1):55?71, 2009.? pages 2, 5, 23, 24, 52J. Guay, D. Boisclair, D. Rioux, M. Leclerc, M. Lapointe, and P. Legendre.Development and validation of numerical habitat models for juveniles ofAtlantic salmon (Salmo salar). Canadian Journal of Fisheries and AquaticSciences, 57(10):2065?2075, 2000. ? pages 4T. Hardy. The future of habitat modeling and instream flow assessmenttechniques. Regulated Rivers: Research & Management, 14(5):405?420, 1998.? pages 1, 2, 4, 6, 22, 51, 52M. Hassan and M. Church. Sensitivity of bed load transport in Harris Creek:Seasonal and spatial variation over a cobble-gravel bar. Water ResourcesResearch, 37(3):813?825, 2001. ? pages 10, 13, 1790T. Hatfield and J. Bruce. Predicting salmonid habitat?flow relationships forstreams from western North America. North American Journal of FisheriesManagement, 20(4):1005?1015, 2000. ? pages 1, 2, 3, 4, 5, 6, 22, 23, 47, 50,52, 84, 86D. Hogan and M. Church. Hydraulic geometry in small, coastal streams: progresstoward quantification of salmonid habitat. Canadian Journal of Fisheries andAquatic Sciences, 46(5):844?852, 1989. ? pages 3, 4H. Q. Huang, H. H. Chang, and G. C. Nanson. Minimum energy as the generalform of critical flow and maximum flow efficiency and for explaining variationsin river channel pattern. Water Resources Research, 40(4), 2004. ? pages 53,54I. Jowett. Instream flow methods: a comparison of approaches. Regulated Rivers:Research & Management, 13(2):115?127, 1997. ? pages 1, 2, 3, 4, 22, 51M. Kirkby. Maximum sediment efficiency as a criterion for alluvial channels.University of Leeds, School of Geography, 1976. ? pages 53J. Laliberte, J. Post, and J. Rosenfeld. Hydraulic geometry and longitudinalpatterns of habitat quantity and quality for rainbow trout (Oncorhynchusmykiss). River Research and Applications, 2013. ? pages 87N. Lamouroux. Depth probability distributions in stream reaches. Journal ofHydraulic Engineering, 124:224?227, 1998. ? pages 2, 6, 7, 8, 23, 26, 29, 30,32, 33, 36, 40, 46, 48N. Lamouroux and H. Capra. Simple predictions of instream habitat modeloutputs for target fish populations. Freshwater Biology, 47(8):1543?1556,2002. ? pages 5, 6, 7N. Lamouroux and I. Jowett. Generalized instream habitat models. CanadianJournal of Fisheries and Aquatic Sciences, 62(1):7?14, 2005. ? pages 5, 6, 7,22, 51, 52N. Lamouroux and Y. Souchon. Simple predictions of instream habitat modeloutputs for fish habitat guilds in large streams. Freshwater Biology, 47(8):1531?1542, 2002. ? pages 5, 52N. Lamouroux, Y. Souchon, and E. Herouin. Predicting velocity frequencydistributions in stream reaches. Water Resources Research, 31(9):2367?2375,1995. ? pages 2, 6, 7, 8, 23, 25, 30, 31, 35, 4691N. Lamouroux, H. Capra, and M. Pouilly. Predicting habitat suitability for loticfish: linking statistical hydraulic models with multivariate habitat use models.Regulated Rivers: Research & Management, 14(1):1?11, 1998. ? pages 4J. Leach and R. Moore. Above-stream microclimate and stream surface energyexchanges in a wildfire-disturbed riparian zone. Hydrological Processes, 24(17):2369?2381, 2010. ? pages 18, 20J. Leach and R. Moore. Stream temperature dynamics in twohydrogeomorphically distinct reaches. Hydrological Processes, 25(5):679?690,2011. ? pages 20M. Leclerc, A. Boudreault, T. Bechara, and G. Corfa. Two-dimensionalhydrodynamic modeling: a neglected tool in the instream flow incrementalmethodology. Transactions of the American Fisheries Society, 124(5):645?662,1995. ? pages 5B. Lee Lamb, C. Sabaton, and Y. Souchon. Use of the instream flow incrementalmethodology: introduction to the special issue. Hydroe?cologie Applique?e, 14:1?7, 2004. ? pages 4L. B. Leopold and T. Maddock. The hydraulic geometry of stream channels andsome physiographic implications. US Geological Survey Professional Paper,252, 1953. ? pages 3MFLNR. Climate change. Technical report, Ministry of Forest, Lands, andNatural Resources - Government of British Columbia, 2009. ? pages 66, 67D. Montgomery and J. Buffington. Channel-reach morphology in mountaindrainage basins. Geological Society of America Bulletin, 109(5):596?611,1997. ? pages 12, 20, 57G. C. Nanson and H. Q. Huang. Least action principle, equilibrium states,iterative adjustment and the stability of alluvial channels. Earth SurfaceProcesses and Landforms, 33(6):923?942, 2008. ? pages 53P. Parasiewicz and M. Dunbar. Physical habitat modelling for fish-a developingapproach. Archiv fu?r Hydrobiologie. Supplementband. Large rivers, 12(2-4):239?268, 2001. ? pages 7P. Parasiewicz and J. Walker. Comparison of MesoHABSIM with twomicrohabitat models (PHABSIM and HARPHA). River Research andApplications, 23(8):904?923, 2007. ? pages 4, 592E. Petticrew, P. Owens, and T. Giles. Wildfire effects on the quantity andcomposition of suspended and gravel-stored sediments. Water, Air, & SoilPollution: Focus, 6(5):647?656, 2006. ? pages 20J. Phillips, B. Eaton, et al. Detecting the timing of morphologic change usingstage-discharge regressions: A case study at Fishtrap Creek, British Columbia,Canada. Canadian Water Resources Journal, 34(3):285?300, 2009. ? pages20QLiner. Adcp technical user manual, 2005. ? pages 16D. Reiser, T. Wesche, and C. Estes. Status of instream flow legislation andpractices in North America. Fisheries, 14(2):22?29, 1989. ? pages 3S. P. Rice, M. Church, C. L. Wooldridge, and E. J. Hickin. Morphology andevolution of bars in a wandering gravel-bed river; lower Fraser River, BritishColumbia, Canada. Sedimentology, 56(3):709?736, 2009. ? pages 57J. Rosenfeld, K. Campbell, E. Leung, J. Bernhardt, and J. Post. Habitat effects ondepth and velocity frequency distributions: Implications for modellinghydraulic variation and fish habitat suitability in streams. Geomorphology, 130:127?135, 2011. ? pages 7, 8, 86J. S. Rosenfeld, J. Post, G. Robins, and T. Hatfield. Hydraulic geometry as aphysical template for the river continuum: application to optimal flows andlongitudinal trends in salmonid habitat. Canadian Journal of Fisheries andAquatic Sciences, 64(5):755?767, 2007. ? pages 87E. Saraeva and T. B. Hardy. Prediction of fisheries physical habitat values basedon hydraulic geometry and frequency distributions of depth and velocity.International Journal of River Basin Management, 7(1):31?41, 2009a. ?pages 1, 5, 7, 22, 23, 26, 30, 31, 40, 46, 47, 48, 49, 51, 52, 55, 84, 86, 87E. Saraeva and T. B. Hardy. Extrapolation of site-specific weighted usable areacurves and instream flow requirements to unmeasured streams within theNooksack watershed in support of strategic watershed planning. InternationalJournal of River Basin Management, 7(1):91?103, 2009b. ? pages 2, 5, 23, 52S. Schweizer, M. E. Borsuk, I. Jowett, and P. Reichert. Predicting joint frequencydistributions of depth and velocity for instream habitat assessment. RiverResearch and Applications, 23(3):287?302, 2007. ? pages 2, 6, 7, 8, 23, 26,29, 30, 31, 32, 33, 34, 35, 36, 38, 40, 41, 42, 46, 47, 52, 55, 59, 62, 63, 80, 83,8493SonTek. Flowtracker Handheld ADV Technical Manual, July 2007. ? pages 14C. B. Stalnaker. Low flow as a limiting factor in warmwater streams. US Fish andWildlife Service, 1979. ? pages 66P. Steffler and J. Blackburn. Two-dimensional depth averaged model of riverhydrodynamics and fish habitat. River2D users manual, University of Alberta,Canada, 2002. ? pages 5, 24M. J. Stewardson and T. A. McMahon. A stochastic model of hydraulic variationswithin stream channels. Water Resources Research, 38(1):1007, 2002. ?pages 38D. Tennant. Instream flow regimens for fish, wildlife, recreation and relatedenvironmental resources. Fisheries, 1(4):6?10, 1976. ? pages 3, 67R. E. Tharme. A global perspective on environmental flow assessment: emergingtrends in the development and application of environmental flow methodologiesfor rivers. River research and applications, 19(5-6):397?441, 2003. ? pages 1,2, 3, 4, 22, 51, 86W. R. White, R. Bettess, and E. Paris. Analytical approach to river regime.Journal of the Hydraulics Division, 108(10):1179?1193, 1982. ? pages 53C. Willmott. Some comments on the evaluation of model performance. Bulletinof the American Meteorological Society, 63:1309?1369, 1982. ? pages 28C. Willmott, S. Ackleson, R. Davis, J. Feddema, K. Klink, D. Legates,J. ODonnell, and C. Rowe. Statistics for the evaluation and comparison ofmodels. Journal of geophysical Research, 90(C5):8995?9005, 1985. ? pages28M. G. Wolman. A method of sampling coarse river-bed material. AmericanGeophysical Union, 1954. ? pages 1794Appendix AHabitat Suitability Indices95Figure A.1: USGS depth and velocity HSI for adult rainbow troutFigure A.2: USGS depth and velocity HSI for juvenile rainbow trout96Figure A.3: USGS depth and velocity HSI for spawning rainbow troutFigure A.4: USGS depth and velocity HSI for adult smallmouth bass97Figure A.5: USGS depth and velocity HSI for adult longnose dace98Appendix BRelative depth and velocitydistributions - Harris Creek99Figure B.1: Relative depth and velocity distributions for Harris Creek at Q= 1.00 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.2: Relative depth and velocity distributions for Harris Creek at Q= 1.23 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.100Figure B.3: Relative depth and velocity distributions for Harris Creek at Q= 1.32 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.4: Relative depth and velocity distributions for Harris Creek at Q= 1.61 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.101Figure B.5: Relative depth and velocity distributions for Harris Creek at Q= 2.29 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.6: Relative depth and velocity distributions for Harris Creek at Q= 2.61 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.102Figure B.7: Relative depth and velocity distributions for Harris Creek at Q= 3.08 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.8: Relative depth and velocity distributions for Harris Creek at Q= 3.51 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.103Figure B.9: Relative depth and velocity distributions for Harris Creek at Q= 3.98 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.10: Relative depth and velocity distributions for Harris Creek at Q= 4.47 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.104Figure B.11: Relative depth and velocity distributions for Harris Creek at Q= 5.55 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.12: Relative depth and velocity distributions for Harris Creek at Q= 7.50 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.105Figure B.13: Relative depth and velocity distributions for Harris Creek at Q =10.00 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.Figure B.14: Relative depth and velocity distributions for Harris Creek at Q =15.00 m3 s?1. The bars represent distributions produced by River2D.The lines are proposed statistical distributions.106Appendix CASHGS habitat indices andsensitivity analyses107Figure C.1: Juvenile rainbow trout WUA and HV at Harris Creek across arange of potential Qb. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure C.2: Spawning rainbow trout WUA and HV at Harris Creek across arange of potential Qb. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.108Figure C.3: Juvenile rainbow trout WUA and HV at Harris Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure C.4: Spawning rainbow trout WUA and HV at Harris Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.109Figure C.5: Sensitivity of juvenile rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of Qb and H.Figure C.6: Sensitivity of spawning rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maximumWUA for a range of Qb and H.110Figure C.7: Juvenile rainbow trout WUA and HV at Fishtrap Creek across arange of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure C.8: Sensitivity of juvenile rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of H at Fishtrap Creek.111Figure C.9: Spawning rainbow trout WUA and HV at Fishtrap Creek acrossa range of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure C.10: Sensitivity of spawning rainbow trout WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of H at Fishtrap Creek.112Figure C.11: Adult smallmouth bass WUA and HV at Fishtrap Creek acrossa range of potential H. The red and blue dash vertical lines representminimum mean monthly flow and mean annual flow respectively.Figure C.12: Sensitivity of adult smallmouth bass WUA at minimum meanmonthly flow (low flow) and mean annual flow as well as the maxi-mum WUA for a range of H at Fishtrap Creek.113

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.24.1-0074274/manifest

Comment

Related Items