"Arts, Faculty of"@en . "Geography, Department of"@en . "DSpace"@en . "UBCV"@en . "West, David"@en . "2014-08-26T20:43:03Z"@en . "2014"@en . "Master of Science - MSc"@en . "University of British Columbia"@en . "The objective of this study was to (1) identify the dominant processes affecting stream temperature and (2) develop and apply a process-based stream temperature model to assess the potential for managing the thermal regime of a regulated river in coastal BC through bank re-forestation and changes to release regime. The study spanned June to September 2013, when temperatures are highest and multiple salmonid species begin their spawning runs. Following June 14, a steady discharge of approximately 2.7 m\u00B3/s was released from the reservoir through a low level outlet. The reservoir developed two-layer stratification and an internal seiche that caused oscillating withdrawal of epilimnetic and hypolimnetic water that was observed to propagate at least 14 km downstream using wavelet analysis.\n\nThe stream temperature model was developed with field measurements of hydrological, meteorological, and geomorphic variables that govern stream heat fluxes. A Lagrangian model was employed to model a 14 km reach of Alouette River beginning at the dam. Using a 10-min time step, an accuracy of 0.54 \u00B0C RMSE was achieved over all predictions, and 0.50 \u00B0C RMSE for daily peaks. Water parcels arriving at the downstream extent near sunset gained most heat through direct and diffuse shortwave radiation, with the greatest cooling effect associated with evaporation, tributary inflows, and bed heat conduction. Parcels arriving near sunrise gained most heat through bed heat conduction, with the greatest cooling effect associated with longwave radiation and tributary inflows.\n\nModelling a bank re-forestation scenario resulted in study reach temperature reductions of less than 0.5 \u00B0C for mean and daily peak temperatures. A lower reservoir outlet withdrawing from the cool hypolimnion resulted in mean daily peak reductions of 4.6 \u00B0C, while raising the outlet into the epilimnion resulted in mean daily peak increases of 0.5 \u00B0C. Increased reservoir discharge caused minor increases of mean and daily peak temperatures of less than 0.15 \u00B0C. When combined, increased discharge amplified the effect of modified release depth. The results of this study stress the importance of basing warming mitigation efforts on a detailed understanding of stream heat fluxes and reservoir stratification regime in regulated systems."@en . "https://circle.library.ubc.ca/rest/handle/2429/50199?expand=metadata"@en . "Modelling the thermal regime of aregulated coastal British Columbiariver and assessing the potential ofwarming mitigation strategiesbyDavid WestB.Sc., Queen\u00E2\u0080\u0099s University, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFMASTER OF SCIENCEinThe Faculty of Graduate and Postdoctoral Studies(Geography)THE UNIVERSITY OF BRITISH COLUMBIA(Vancouver)August 2014c\u00C2\u00A9 David West 2014AbstractThe objective of this study was to (1) identify the dominant processes affecting stream temperatureand (2) develop and apply a process-based stream temperature model to assess the potential formanaging the thermal regime of a regulated river in coastal BC through bank re-forestation andchanges to release regime. The study spanned June to September 2013, when temperatures arehighest and multiple salmonid species begin their spawning runs. Following June 14, a steadydischarge of approximately 2.7 m3/s was released from the reservoir through a low level outlet. Thereservoir developed two-layer stratification and an internal seiche that caused oscillating withdrawalof epilimnetic and hypolimnetic water that was observed to propagate at least 14 km downstreamusing wavelet analysis.The stream temperature model was developed with field measurements of hydrological, me-teorological, and geomorphic variables that govern stream heat fluxes. A Lagrangian model wasemployed to model a 14 km reach of Alouette River beginning at the dam. Using a 10-min timestep, an accuracy of 0.54 \u00E2\u0097\u00A6C RMSE was achieved over all predictions, and 0.50 \u00E2\u0097\u00A6C RMSE for dailypeaks. Water parcels arriving at the downstream extent near sunset gained most heat throughdirect and diffuse shortwave radiation, with the greatest cooling effect associated with evapora-tion, tributary inflows, and bed heat conduction. Parcels arriving near sunrise gained most heatthrough bed heat conduction, with the greatest cooling effect associated with longwave radiationand tributary inflows.Modelling a bank re-forestation scenario resulted in study reach temperature reductions of lessthan 0.5 \u00E2\u0097\u00A6C for mean and daily peak temperatures. A lower reservoir outlet withdrawing fromthe cool hypolimnion resulted in mean daily peak reductions of 4.6 \u00E2\u0097\u00A6C, while raising the outletinto the epilimnion resulted in mean daily peak increases of 0.5 \u00E2\u0097\u00A6C. Increased reservoir dischargecaused minor increases of mean and daily peak temperatures of less than 0.15 \u00E2\u0097\u00A6C. When combined,increased discharge amplified the effect of modified release depth. The results of this study stressthe importance of basing warming mitigation efforts on a detailed understanding of stream heatfluxes and reservoir stratification regime in regulated systems.iiPrefaceThis dissertation is original, unpublished, independent work by the author, D. West. Assistancewith developing the stream temperature model and bed heat conduction model was provided byR. D. Moore and J. A. Leach.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Motivation for the study . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Effects of elevated stream temperature on salmonids . . . . . . . . . . . . . . . . . . 21.3 Thermal regime of regulated rivers . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.3.1 Effects of impoundment on release temperature . . . . . . . . . . . . . . . . 31.3.2 Effects of impoundment on downstream temperature . . . . . . . . . . . . . 41.4 Climatic change effects on stream temperature . . . . . . . . . . . . . . . . . . . . . 51.5 Strategies to mitigate elevated stream temperature . . . . . . . . . . . . . . . . . . . 61.6 Research questions and thesis structure . . . . . . . . . . . . . . . . . . . . . . . . . 92 Study site and data collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.1 Site description . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.2 Field measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.1 Stream and bed temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.2.2 Streamflow and stream-subsurface exchanges . . . . . . . . . . . . . . . . . . 162.2.3 Hydraulic geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.4 Meteorology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.5 Riparian canopy characteristics . . . . . . . . . . . . . . . . . . . . . . . . . 21ivTable of Contents3 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.1 Time series analysis of stream temperature . . . . . . . . . . . . . . . . . . . . . . . 223.1.1 Seiching signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223.1.2 Wavelet analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 233.1.3 Cross-correlation analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 243.2 Stream temperature model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2.1 Model structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 253.2.2 Above-stream microclimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 263.2.3 Turbulent exchanges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 273.2.4 Net radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 283.2.5 Tributary flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.2.6 Mainstem flow calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.2.7 Groundwater distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.2.8 Hyporheic exchange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333.2.9 Inflow temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.2.10 Bed heat conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343.2.11 Uncertainty/sensitivity analysis . . . . . . . . . . . . . . . . . . . . . . . . . 353.3 Management scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 374 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.1 Overview of the study period . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.2 Time series analysis of stream temperature . . . . . . . . . . . . . . . . . . . . . . . 444.2.1 Overview of stream temperature time series . . . . . . . . . . . . . . . . . . 444.2.2 Seiching signal . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 504.2.3 Wavelet analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 514.2.4 Cross-correlation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 594.3 Field measurements for model parameterization . . . . . . . . . . . . . . . . . . . . 654.3.1 Reach-scale hydrology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 654.3.2 Shade measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.3.3 Channel geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.3.4 Bed and bank temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . 734.3.5 Tributary temperatures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 784.3.6 Electrical conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 794.3.7 Above-stream microclimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.4 Data processing for stream temperature modelling . . . . . . . . . . . . . . . . . . . 834.4.1 Flow distribution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.4.2 Stream surface radiation modelling . . . . . . . . . . . . . . . . . . . . . . . 87vTable of Contents4.4.3 Above-stream microclimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 914.4.4 Bed heat conduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 924.4.5 Hydraulic geometry with increased flow . . . . . . . . . . . . . . . . . . . . . 954.5 Stream temperature model evaluation . . . . . . . . . . . . . . . . . . . . . . . . . . 964.6 Modelled heat fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014.7 Management scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1045 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1125.1 Stream temperature patterns . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1125.2 Model components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1145.2.1 Radiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1145.2.2 Mainstem flow and inflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1155.2.3 Turbulent exchanges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1165.2.4 Bed heat fluxes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1165.2.5 Inflow and groundwater temperature . . . . . . . . . . . . . . . . . . . . . . 1175.2.6 Hydraulic geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185.3 Model performance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185.3.1 Model performance compared to similar studies . . . . . . . . . . . . . . . . 1185.3.2 Error trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185.4 Management scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1205.4.1 Bank re-forestation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1205.4.2 Reservoir release regime . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1215.4.3 Other management options . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1225.5 Potential effects of climate change . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1225.6 Implications for aquatic habitat . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1226 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1246.1 Key findings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1246.2 Directions for future research . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128viList of Tables1.1 Studies applying process based models to predict the effects of climate change. . . . 82.1 Equipment used for data collection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.2 Shade transect measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213.1 Summary of inputs used in model uncertainty analysis runs. . . . . . . . . . . . . . . 364.1 Historical air temperature and precipitation recorded at a BC Hydro managed weatherstation on Alouette Dam. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 404.2 Bed and bank piezometer head gradients. . . . . . . . . . . . . . . . . . . . . . . . . 704.3 Shade transect measurements for each shade reach. . . . . . . . . . . . . . . . . . . . 714.4 Summary of observed hydraulic geometry in each shade reach. . . . . . . . . . . . . 724.5 Bank piezometer temperature measurements. . . . . . . . . . . . . . . . . . . . . . . 774.6 Summary of regressions relating above-stream to open-site climate variables. . . . . 924.7 Summary of model error for runs including the full study reach, the upper half, andthe lower half. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1014.8 Model results for management scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . 111viiList of Figures2.1 Location of the study reach within southwest British Columbia. . . . . . . . . . . . . 112.2 Location of mainstem Tidbit temperature loggers. . . . . . . . . . . . . . . . . . . . 142.3 Shade transects and subsurface monitoring site locations. . . . . . . . . . . . . . . . 162.4 Location of gauged tributaries and mainstem flow measurement stations. . . . . . . . 172.5 Above-stream microclimate station locations. . . . . . . . . . . . . . . . . . . . . . . 203.1 Tributary subcatchments delineated with SAGA GIS. . . . . . . . . . . . . . . . . . 324.1 Climate variables measured at Alouette Dam over the study period. . . . . . . . . . 414.2 Open site radiation calculated at the open site climate station on Alouette Dam. . . 424.3 Historical Alouette flows (1998-2012) compared to 2013 study period flows. . . . . . 434.4 Stream temperature time series June 6-13. . . . . . . . . . . . . . . . . . . . . . . . . 454.5 Stream temperature time series June 23-30. . . . . . . . . . . . . . . . . . . . . . . . 464.6 Stream temperature time series July 22-30. . . . . . . . . . . . . . . . . . . . . . . . 474.7 Stream temperature time series August 21-29. . . . . . . . . . . . . . . . . . . . . . . 494.8 Alouette Lake temperature variation during the 2013 study period. . . . . . . . . . . 504.9 Wavelet transform power spectra for the reservoir outlet temperature time series. . . 524.10 Wavelet transforms of temperature time series at points of interest along AlouetteRiver, North Alouette River, and Millionaire Creek. . . . . . . . . . . . . . . . . . . 564.11 Cross wavelet transform plots for the reservoir outlet temperature crossed with pointsof interest along Alouette River. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 584.12 Temperature time series T1 filtered with a 9-15 h band-pass filter. . . . . . . . . . . 594.13 Cross-correlation plots for the reservoir outlet temperature crossed with points ofinterest along Alouette and North Alouette River. . . . . . . . . . . . . . . . . . . . 614.14 Time lags associated with MPC from cross-correlation analysis compared to traveltimes from estimated velocity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 624.15 Cross-correlation of river stage in the tidally influenced region with temperaturesalong Alouette and North Alouette Rivers. . . . . . . . . . . . . . . . . . . . . . . . . 644.16 Streamflow time series at boundaries of study reach. . . . . . . . . . . . . . . . . . . 654.17 Mainstem streamflow measurements taken between the dam and WSC gauge. . . . . 66viiiList of Figures4.18 Mainstem streamflow as a function of distance from Alouette Dam. . . . . . . . . . . 674.19 Mainstem streamflow as a function of drainage area. . . . . . . . . . . . . . . . . . . 684.20 Streamflow measurements at nine tributaries with additional flows predicted fromstage measurements. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694.21 Samples of oblique and hemispherical photographs in different shade reaches. . . . . 714.22 Representative cross sections from shade transects in each shade reach. . . . . . . . . 734.23 Stream bed temperature at subsurface monitoring sites part 1. . . . . . . . . . . . . 744.24 Stream bed temperature at subsurface monitoring sites part 2. . . . . . . . . . . . . 754.25 Bed temperature profiles measured with bed temperature stakes. . . . . . . . . . . . 764.26 Floodplain temperatures at locations where moderate groundwater upwelling oc-curred continuously. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.27 Right bank and left bank tributary temperatures throughout parts of the study period. 794.28 Tributary mean electrical conductivities. . . . . . . . . . . . . . . . . . . . . . . . . . 804.29 Air temperature at above-stream sites compared to open site. . . . . . . . . . . . . . 814.30 Relative humidity at above-stream sites compared to open site. . . . . . . . . . . . . 824.31 Wind speed at above stream sites compared to open site. . . . . . . . . . . . . . . . 834.32 Streamflow at the WSC gauge, with raw data and loess smoothed data to removehourly noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.33 Proportion of total study reach inflows at each flow gauging site on each flow mea-surement day. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854.34 Tributary streamflows as a function of streamflow at WSC gauge. . . . . . . . . . . . 864.35 Overland and groundwater inflow distribution over the full study reach. . . . . . . . 874.36 Net radiation measured with the net radiometer and modelled with hemisphericalphotos taken at the same locations as the net radiometer. . . . . . . . . . . . . . . . 884.37 Modelled incident shortwave radiation at select shade transects on a clear sky day. . 894.38 Reach-averaged direct shortwave radiation on a clear sky day and partially cloudyday. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 904.39 Modelled sky view factors for each shade transect and view factor calibration fitusing hemispherical modelled sky view factors. . . . . . . . . . . . . . . . . . . . . . 914.40 Linear regression fits of above-stream climate station variables as a function of opensite climate variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 924.41 Modelled bed temperatures at Tidbit T20 during week of MWMT. . . . . . . . . . . 934.42 Modelled bed temperatures at Tidbit T5 during week of MWMT. . . . . . . . . . . . 944.43 Modelled bed heat conduction in each of the four flow reaches during the week ofMWMT. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 954.44 Modelled hydraulic geometry with increasing flows. . . . . . . . . . . . . . . . . . . . 964.45 Model uncertainty analysis runs during the week of MWMT. . . . . . . . . . . . . . 98ixList of Figures4.46 Modelled stream temperature for full study period. . . . . . . . . . . . . . . . . . . . 994.47 Modelled stream temperature up to midpoint of study reach. . . . . . . . . . . . . . 1004.48 Modelled stream temperature from midpoint to end of study reach. . . . . . . . . . . 1004.49 Net heat exchanged by each modelled heat flux as a parcel of water flows throughthe study reach. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1024.50 Stream heat fluxes as a percentage of total heat exchange. . . . . . . . . . . . . . . . 1034.51 Daily mean temperatures modelled at the WSC gauge with re-forested banks andchanges to withdrawal elevation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1054.52 Daily mean temperatures modelled at the WSC gauge with changes to dam dischargeand combinations of discharge and release elevation changes. . . . . . . . . . . . . . 1064.53 Daily maximum temperatures modelled at the WSC gauge with re-forested banksand changes to withdrawal elevation. . . . . . . . . . . . . . . . . . . . . . . . . . . . 1074.54 Daily maximum temperatures modelled at the WSC gauge with changes to damdischarge and combinations of discharge and release elevation changes. . . . . . . . . 1084.55 Modelled temperatures at the WSC gauge during the week of MWMT with re-forested banks and changes to withdrawal elevation. . . . . . . . . . . . . . . . . . . 1094.56 Modelled temperatures at the WSC gauge during the week of MWMT with changesto dam discharge and combinations of discharge and release elevation changes. . . . 110xAcknowledgementsFirstly, I would like to thank my supervisor, Dan Moore, for his technical support, attentiveness,thorough explanations, and guidance. I am also grateful to my supervisory committee members,Brett Eaton and Gregory Lawrence, for ideas and support. I am indebted to Jason Leach, JoelTrubilowicz, and Lawrence Bird for technical support in the office, and to Justin Knudson, AlistairDavis, and Brent Wilson for support in the field. The level of detail would not have been possiblewithout the data generously provided by Shannon Harris, Jeff Greenbank, Georg Jost, and JanetWong. Additional ideas and support were provided by UBC faculty members including JohnRichardson, Scott Hinch, and Roger Pieters. My parents and family have been supportive in allof my endeavors. Finally, Julie has made this journey a worthwhile and enjoyable one. Thisresearch was funded by a Natural Sciences and Engineering Research Council Discovery Grant toR.D. Moore, a Teaching Assistantship from the UBC Department of Geography, and a GraduateSupport Initiative from the UBC Faculty of Graduate and Postdoctoral Studies.xiChapter 1Introduction1.1 Motivation for the studyThe thermal regime of a stream influences habitat suitability for aquatic species, rates of growth anddevelopment of aquatic organisms, and water quality parameters including dissolved oxygen andtoxicity of contaminants (Poole and Berman, 2001; Caissie, 2006; Ficke et al., 2007). Commerciallyand ecologically valuable species such as salmonids have narrow thermal tolerances that, whenexceeded, can cause severe health impacts, prevent spawning, or cause mortality (Coutant, 1999;Caissie, 2006).Stream thermal regimes reflect the interaction between hydrologic and meteorological processes,both of which can be influenced by climatic variability and change, and by human activities thatmodify the water and energy budgets of a stream reach. Common sources of thermal disturbanceinclude loss of riparian vegetation (e.g., associated with forestry) (Moore et al., 2005a; Gomi et al.,2006; Leach and Moore, 2010), impermeable urbanized catchments (Krause et al., 2004; Nelson andPalmer, 2007), flow withdrawals (Hockey et al., 1982; Sinokrot and Gulliver, 2000; Risley et al.,2010), heated industrial effluent (Coutant, 1999), and impoundments for hydroelectricity generationand water supply (Sinokrot et al., 1995; Webb and Walling, 1997; Gu et al., 1998; Sinokrot andGulliver, 2000; Hamblin and McAdam, 2003; Gooseff et al., 2005; Yearsley, 2009; Risley et al.,2010; Null et al., 2013). Whereas the first four of these stressors generally cause elevated summertemperatures, the effects of impoundments can be complicated, depending on the thermal structureof the reservoir and the nature of the release structure (Olden and Naiman, 2010).Increased stream temperatures associated with thermal disturbances can degrade habitat suit-ability for cold- and cool-water species, and in many areas of North America, once-productivestreams have become unsuitable for spawning (Sinokrot et al., 1995; McCullough, 1999; Coutant,1999). Rising air temperature and other climatic changes are projected to cause stream warming,causing a substantial loss of habitat in streams already approaching thermal thresholds (Cristeaand Burges, 2010). Many studies have focused on the effects of anthropogenic stressors on streamtemperature including climate change (Pilgrim et al., 1998; Mohseni et al., 1999; Yearsley, 2009;Isaak, 2011; Punzet et al., 2012; Isaak et al., 2012), but only a few studies have assessed the poten-tial for management intervention (Sinokrot et al., 1995; Gooseff et al., 2005; Cristea and Burges,2010; Hrachowitz et al., 2010).11.2. Effects of elevated stream temperature on salmonidsThis study focused on the thermal regime of Alouette River, a regulated river in coastal BritishColumbia, Canada, in which stream temperature approaches or occasionally exceeds thermal thresh-olds for salmonids that reside or spawn in the river (Caryula and Groves, 2009; Greenbank, 2009).The main objectives were (a) to contribute to the process-based understanding of the thermalregime of regulated rivers and (b) to assess the extent to which mitigation efforts could reduce thefrequency and magnitude of thermal threshold exceedance for salmonids. The study focused on thelow flow summer period when temperatures are highest and various salmonid species are beginningtheir spawning runs.The remainder of this chapter provides brief literature reviews to provide context for the specificresearch questions addressed by this study. The reviews focus on the following topics: the effectsof high water temperature on salmonids, thermal regimes of regulated rivers, the effects of climatechange on river thermal regime, and potential strategies to mitigate thermal disturbances. Thechapter ends with a statement of the research questions and the organization of the remainder ofthe thesis.1.2 Effects of elevated stream temperature on salmonidsBiological and ecological studies have quantified the effects of elevated stream temperature onaquatic organism survival, health, and recruitment (Caissie, 2006). For the purpose of this review,discussion will be limited to the effects on salmonids, although stream temperature also has astrong influence on the population dynamics of invertebrates and other processes that ultimatelyinfluence the food supply for salmonids (Ficke et al., 2007). Anadromous fish species, includingsalmonids, have been referred to as a \u00E2\u0080\u009Ckeystone species\u00E2\u0080\u009D in ecological literature because they makea critical contribution to terrestrial and lacustrine ecosystems through chemical nutrients and as afood source sustaining predator species (Willson and Halupka, 1995).Anadromous salmonids have developed a variety of life-cycle strategies resulting in differ-ing spawning run and seaward migration timing across species and within separate populations(Coutant, 1999). Many salmonid populations complete spawning runs during autumn, when tem-peratures typically recede to a non-critical level. However, some populations demonstrate con-trasting life-cycles, such as spring-run chinook salmon that spend the entire summer in freshwaterstreams, hiding in cold water refugia such as deep pools before spawning in autumn (Coutant,1999). Some sockeye populations complete spawning runs during summer, when the threat ofthermal stress is the highest (Cooke et al., 2004).The effects of water temperature on salmonids during warmer seasons have garnered the mostresearch attention when temperatures approach the upper threshold for migration success (Morrisonet al., 2002; Cristea and Burges, 2010). The most vulnerable species are likely the early runpopulations (Coutant, 1999), which can begin spawning runs as early as June (Cooke et al., 2004),21.3. Thermal regime of regulated riverswhen a lack of precipitation results in shallower, slower flows that are more vulnerable to heating(Nelson and Palmer, 2007; Webb et al., 2008).Temperatures that are only modestly above average, yet well below acute lethal levels, canhave a negative effect on spawning success (Crossin et al., 2008). Acute lethal conditions occurbetween 21 and 28 \u00E2\u0097\u00A6C for trout and salmon. However, it has been found that sockeye salmon withchronic exposure to temperatures >18 \u00E2\u0097\u00A6C experience 20% to 100% mortality en route to spawning(Crossin et al., 2008). Higher temperatures have also been connected with increased pathogenicity ofbacterial diseases, infections, and parasites (McCullough, 1999; Crossin et al., 2008) with optimumgrowth conditions above 15 \u00E2\u0097\u00A6C (McCullough, 1999).1.3 Thermal regime of regulated riversReservoirs have been found to impact downstream thermal regimes with varying degrees of intensity(Olden and Naiman, 2010). The effect depends on the characteristics of the stream and watershed,as well as the size and operation of the reservoir.1.3.1 Effects of impoundment on release temperatureLakes and reservoirs in temperate regions typically stratify during summer. Stratified lakes havea deep cooler layer (hypolimnion) that is less subject to the effects of surface energy fluxes thataffect the upper layer (epilimnion), especially during summer (Elo et al., 1998). The boundarybetween the epilimnion and the hypolimnion is a layer with strong vertical temperature gradients(the thermocline). Reservoir stratification timing and temperature of layers is dependent on mete-orological and hydrological variables as well as reservoir operation (Webb and Walling, 1997; Nullet al., 2013). For example, a lower release rate can cause an increased flow retention time andassociated increase in duration of surface heat absorption (Hamblin and McAdam, 2003; Olden andNaiman, 2010). In temperate regions, the depth of the thermocline typically increases through thesummer as the surface layer warms, but can be influenced by synoptic-scale weather systems withstrong winds, which can overcome the stability of the reservoir and drive vertical mixing.The temperature of discharge from stratified reservoirs depends on the depth of the outletand release rate. Most large dams have hypolimnetic releases designed to provide cooler waterfor salmonid species (Sinokrot et al., 1995; Webb and Walling, 1997; Yearsley, 2009; Olden andNaiman, 2010). Other smaller reservoirs can have epilimnetic releases when the depth is not greatenough for stratification to form (Gooseff et al., 2005), or a hypolimnetic release was not designed(Lessard and Hayes, 2003). In the late summer, reservoir levels have often declined, which resultsin a thinner hypolimnion and thermocline at a lower elevation (Null et al., 2013). Withdrawal layerthickness is also affected by discharge rate, such that a higher discharge rate is capable of extractinga thicker portion of the water column (Spigel and Farrant, 1984; Hamblin and McAdam, 2003).31.3. Thermal regime of regulated riversThis can result in the withdrawal of more hypolimnetic water or epilimnetic water with increaseddischarge depending on whether a reservoir\u00E2\u0080\u0099s outlet is above or below the thermocline (Hamblinand McAdam, 2003). The effect of hydroelectric dam construction on stream hydrographs in thePacific coast of North America can be an inversion of the seasonal pattern, in which spring peakflows are reduced and winter flows are increased (Hamblin and McAdam, 2003; Null et al., 2013).Management options for controlling release temperatures generally consist of selective with-drawal from multiple outlets at different depths (Webb and Walling, 1997; Yearsley, 2009) or me-chanical mixing of the reservoir near the outlet to break up stratification (Olden and Naiman, 2010).Additionally, promoting a springflow component through the dam to the receiving stream has beenassessed to be an effective cooling method (Webb and Walling, 1997; Sinokrot and Gulliver, 2000)1.3.2 Effects of impoundment on downstream temperatureThe effect of reservoir release temperature and discharge rate on receiving streams is often anunnaturally high or low tail water temperature that progressively recovers downstream towardsthe natural thermal regime (Olden and Naiman, 2010). During summer stratified conditions, theeffect of reservoir release temperature on downstream temperatures can be a warming effect withan epilimnetic release (Webb and Walling, 1997; Lessard and Hayes, 2003; Gooseff et al., 2005),or a cooling effect with hypolimnetic release (Sinokrot et al., 1995; Yearsley, 2009; Risley et al.,2010). The effect is also dependent on discharge rate and timing, where an unnatural hydrographcan cause altered temperature patterns compared to the typical diurnal oscillation, and a largedegree of inter-annual variability (Webb and Walling, 1997; Risley et al., 2010)). A shift in thetiming of peak temperatures often occurs from summer to mid-autumn associated with reservoirlevel and stratification dynamics (Risley et al., 2010; Null et al., 2013). Reductions in streamfloware associated with a reduction in thermal inertia, which can lead to greater summer diurnaltemperature variations associated with a greater influence of surface and subsurface heat exchanges(Hockey et al., 1982; Gu et al., 1998; Sinokrot and Gulliver, 2000; Risley et al., 2010). Polehn andKinsel (1997) and Lowney (2000) observed locations of minimal and maximal diurnal variationdownstream of a reservoir with constant discharge and release temperature. The locations ofmaximal diurnal variation corresponded to 12-h travel time downstream, since a parcel of waterleaving at daybreak will experience 12 h of daytime (warming) fluxes, and a parcel leaving at sunsetwill experience 12 h of nighttime (cooling) fluxes.The distance required for recovery to pre-dam conditions or adjusted equilibrium temperaturebased on the altered flow regime is dependent on climatic, hydrological, geomorphological, andphysiographic parameters of the watershed. The equilibrium temperature of a stream is the tem-perature at which the net rate of surface heat exchange is equal to zero (Edinger et al., 1968).Downstream of a reservoir, stream temperature will progress towards the equilibrium temperature,41.4. Climatic change effects on stream temperatureassuming a minor influence from advective heat exchanges (Moore et al., 2005a). Recovery distanceto pre-dam conditions has typically been found to be over 40 km, with some streams taking as muchas 930 km (Olden and Naiman, 2010). Sinokrot et al. (1995) found that on four streams with meanannual streamflows between 45-517 m3/s, temperatures could increase by 5 \u00E2\u0097\u00A6C or decrease by 10\u00E2\u0097\u00A6C at a location 48 km downstream of a reservoir, and not experience full recovery to pre-damtemperatures.The dominant heat fluxes affecting thermal recovery will vary depending on the size of a stream,since smaller headwater streams often have a higher influence from advective processes includinghyporheic exchange, groundwater inflows, and snowmelt, while larger streams have a greater in-fluence from surface energy exchanges due to increased width to depth ratio and lower velocity(Moore et al., 2005a; Null et al., 2013). Width to depth ratio was found to strongly influence therecovery of Platte River temperatures downstream of a dam in Nebraska, USA, which could beremedied with flow augmentation (Sinokrot and Gulliver, 2000). Risley et al. (2010) found thatgroundwater withdrawal from the riparian area downstream of a reservoir caused warming duringsummer months, which highlights the potential importance of groundwater as a cooling mechanism.Null et al. (2013) found that cool tributaries could reduce the temperature of regulated streamsin the Sierra Nevada mountain region of California, US. Olden and Naiman (2010) also describedmultiple studies where tributaries were influential in thermal recovery.1.4 Climatic change effects on stream temperatureOver the last few decades, global temperatures have been rising, and south coastal British Columbiahas experienced increases in both minimum and maximum daily air temperatures (Rodenhuis et al.,2007). Based on climate models, it has been projected that this region will experience further airtemperature increases of 1.8 - 2.6 \u00E2\u0097\u00A6C (Pike et al., 2010; Shrestha et al., 2012) between 2050 and2099 (Rodenhuis et al., 2007; Pike et al., 2010). Although projections derived from general circu-lation models (GCMs) or regional circulation models (RCMs) often disagree in the magnitude andsometimes direction of change, the majority make similar predictions for changes in air temperaturein the study region (Rodenhuis et al., 2007; Murdock and Spittlehouse, 2011). There is greaterdisagreement among model projections for precipitation, but the majority suggest that summerswill be drier than at present (Rodenhuis et al., 2007; Murdock and Spittlehouse, 2011).Higher air temperatures would be associated with greater surface energy inputs via sensibleheat exchange and incident longwave radiation, which would tend to promote warming, all otherfactors being held equal. In addition, the climatic changes projected for the south coastal region ofBritish Columbia should result in reduced summer discharge in nival, hybrid and pluvial catchmentstypical of those in the region (Pike et al., 2010) via (a) an earlier and reduced snowmelt freshet,(b) a reduction in summer rainfall runoff, and (c) increased evapotranspiration. Reduced summer51.5. Strategies to mitigate elevated stream temperaturestreamflow would be associated with shallower flow in unregulated streams, and thus a greaterthermal response to energy inputs at the stream surface (Hockey et al., 1982; Stefan and Sinokrot,1993; Sinokrot and Gulliver, 2000). Groundwater temperature is also expected to increase, as ittends to be related to mean annual air temperature (Sinokrot et al., 1995). Groundwater is typicallycooler than stream water during summer days, and thus has a cooling effect at those times (Risleyet al., 2010). Warming of groundwater would thus be a secondary, but possibly still significant,driver of elevated summer stream temperature.For a regulated river, climatic change could influence stream temperatures through (a) changesin release temperatures and (b) changes to energy and water exchanges downstream of the dam. Inrelation to release temperatures, the position of the thermocline has been projected to deepen as aresult of climate change (Schindler et al., 1996; Johnson et al., 2004). If the thermocline droppedbelow the level of the intake for flow releases, the consequence would be a significant increase in thetemperature of the flow releases. For a regulated river such as the Alouette, releases are managedand thus not directly influenced by climatic change. However, the projected climatic changes shouldresult in reduced tributary inflow, warmer groundwater discharge, and higher surface energy inputsvia sensible heat and longwave radiation. Considering these points, Alouette River temperaturesbelow the dam are likely to be further elevated under future climatic conditions, with increasingrisks of thermal tolerance levels for salmonids.1.5 Strategies to mitigate elevated stream temperatureOne potential strategy for mitigating the effects of climatic change on the thermal regime of aregulated river is to increase flow releases. However, the efficacy of this strategy depends on thetemperature of the flow releases, which depends, in turn, on the thermal structure of the reservoirand the depth from which releases are drawn. One approach to mitigating this effect would be toraise the reservoir level to allow cooler water to be withdrawn from low level outlets. However,deepening a reservoir can result in increased overbank extension, where reservoir water surfacebacks up onto the floodplain of the inflow source river, resulting in warming due to a larger surfacearea with low depth (Coutant, 1999). Reservoirs can also be used to re-establish cool season peakflows through controlled releases, which can restore aquifer and floodplain storage of cool waterthat discharges throughout the warm summer (Coutant, 1999).Another potential strategy for mitigating stream warming is to increase shading by riparianforest (Cristea and Burges, 2010; Lawrence et al., 2014). This strategy would be most effective forsmall to intermediate streams, for which riparian forest can provide substantial shade. The portionof the stream that is shaded depends not only on stream width and the height, density and speciesof riparian trees, but also on stream orientation (Chen et al., 1998a,b; Cristea and Burges, 2010;Li et al., 2012). For example, a stream oriented north-south could receive a substantial amount of61.5. Strategies to mitigate elevated stream temperaturesolar radiation around noon despite the presence of riparian forest.The most rigorous approach to assessing the potential effectiveness of mitigation strategies isto apply a physically based deterministic model. These models can be developed using one of twoframes of reference: (1) modelling the change in stream temperature at specific locations along thestream, known as an Eulerian frame of reference (Sinokrot and Stefan, 1993; Sinokrot et al., 1995;Gooseff et al., 2005; Cristea and Burges, 2010) and (2) modelling the temperature of a movingparcel of water as it flows downstream, known as a Lagrangian modelling (Hockey et al., 1982;Polehn and Kinsel, 2000; Leach and Moore, 2011). The Lagrangian approach can provide higherstability and accuracy, while the Eulerian method can be more convenient as it uses a fixed grid(Yearsley, 2009). Models applying an Eulerian frame of reference often require high spatial andtemporal resolution in order to avoid stability issues (Gooseff et al., 2005; Yearsley, 2009). Toovercome some of the shortcomings of each system, hybrid Eulerian-Lagrangian methods have alsobeen employed (Yearsley, 2009).Most stream temperature models are one-dimensional. That is, they assume that temperaturesare uniform across the stream channel and only vary along the channel length. In addition, somemodels can accommodate unsteady flow (i.e., stream discharge varies with time) (e.g., Stefan andSinokrot (1993), while others assume steady flow (e.g., Cristea and Burges (2010)). During extendedperiods of low streamflow, streamflow tends to change slowly through time and can be reasonablyapproximated as steady state.A small number of studies have applied deterministic models to predict the effects of climaticchange and, in some cases, mitigation potential has also been assessed (Table 1.1). Daily meanstream temperature was often selected as the temporal interval, which disregards the potentialfor acute lethal temperatures to occur at the subdiel scale (Gooseff et al., 2005). Cristea andBurges (2010) took extreme temperatures into consideration when assessing the effects of bankre-forestation on daily maximum and hourly temperatures. The degree to which the studies as-sessed the effects of changing reservoir thermal regime and outflow varied despite each of the studystreams having an upstream reservoir. The effects of climate change on a reservoir were predictedby Sinokrot et al. (1995) and Yearsley (2009), who incorporated hydrology and reservoir hydrody-namics models to simulate reservoir thermal regime.71.5.StrategiestomitigateelevatedstreamtemperatureTable 1.1: Studies applying process based models to predict the effects of climate change.Study Study Reservoir effects Model Resolution Calibration Scenarios and averagelocation considered formulation Accuracy (\u00E2\u0097\u00A6C) predicted warming (\u00E2\u0097\u00A6C)Stefan and Sinokrot Minnesota, No Eulerian, 1D, Daily 0.46 2xCO2: 2.4 to 4.71993 USA Unsteady (6 with vegetation loss)Sinokrot et al. Minnesota, Yes Eulerian, 1D, Daily 0.2 to 1.0 2xCO2: 41995 USA UnsteadyMorrison et al. Fraser River, No Eulerian, 1D, Hourly 1.12 2070 to 2099: 1.92002 BC UnsteadyGooseff et al. Montana, No Eulerian, 1D, Hourly 0.118 to 0.229 2xCO2: 28% increase in fish2005 USA Unsteady critical threshold exceedanceCristea & Burges Washington, No Eulerian, 1D, Hourly 0.28 to 0.78 2020s: 1 to 1.2, 2040s: 22009 USA Steady 2080s: 2.5 to 3.6Yearsley Washington & Yes Semi-Lagrangian, Hourly 0.03 to 0.63 2020s: 0.8 to 1.22009 Idaho, USA 1D, Steady 2040s: 1.6 to 281.6. Research questions and thesis structure1.6 Research questions and thesis structureThe general objective of this study is to identify the processes affecting stream temperature onthe regulated Alouette River and use this knowledge to assess the warming mitigation potential ofmitigation strategies. Three sets of research questions assessed in this study are as follows:1. How do downstream temperatures respond to variations in reservoir release temperatureassociated with changing stratification dynamics? How far do these effects propagate downstream,and how do they interact with the effects of stream heat and water exchanges?2. Can a process based stream temperature model developed with summer field measurementsaccurately simulate stream thermal regime downstream of a reservoir? What are the most impor-tant heat fluxes and their driving processes in the Alouette system? Which field measurements anddata sets are most important for simulating the dominant heat fluxes?3. To what extent can high temperatures that degrade fish habitat be mitigated by managementscenarios including re-foresting banks and changes to reservoir release rate and elevation? Whatwould be the effect on downstream temperature of a strictly epilimnetic release throughout thesummer stratified period?The following chapters outline the study site and data collection methodology (Chapter 2),describe the data analysis and model preparation methodology (Chapter 3), summarize the resultsof the stream temperature analysis and modelling exercise (Chapter 4), discuss the results withregards to the research questions outlined above (Chapter 5), and summarize the key findings andrecommendations for future research (Chapter 6).9Chapter 2Study site and data collection2.1 Site descriptionAlouette River is situated in the Pacific Ranges physiographic region of the southern Coast Moun-tains of British Columbia, Canada (Figure 2.1). The valley walls are composed of bedrock withlarge terraces of quaternary sediment including glaciofluvial sandy gravel, glaciolacustrine stoney-silt, and ice-contact deposits (Geological Survey of Canada, 1980). A kame feature on the eastern(left) bank borders the stream from 3 to 6.5 km downstream of the dam, composed of glaciofluvialsandy gravel and gravelly sand outwash (Geological Survey of Canada, 1980). The floodplain iscomposed of glacial outwash sand and gravel (Geological Survey of Canada, 1980).Prior to regulation, Alouette River had a hybrid hydrologic regime dominated by autumn-winter rainfall with a secondary spring snowmelt contribution from higher-elevation portions of thecatchment. The regulated portion of the river drains 36.5 km2 and flows 25 km from the dam toits confluence with Pitt River, a tributary of Fraser River. The regulated reach begins at 116 mabove mean sea level in a steeper forested catchment with a channel gradient of 0.9% (upper reach)until approximately 6.5 km downstream, where the gradient drops to 0.6%, the channel aspectchanges from north-south to east-west, and the river enters a residential area in the city of MapleRidge (lower reach). The study area ends at 14 km below the dam, where the land use switches toprimarily agricultural, the gradient becomes nearly flat, and a tidal influence begins.102.1. Site descriptionFigure 2.1: Location of the study reach within southwest British Columbia.The forested catchment of the upper reach is primarily western hemlock (Tsuga heterophylla)that was planted after clear-cut logging in the 1930s (GVRD, 2004), except for the channel banks,which are dominated by encroaching red alder (Alnus rubra). In the residential area, the riparianforest is primarily deciduous trees including bigleaf maple (Acer macrophyllum) that overhangfurther over the channel.The upper reach channel morphology consists of long sand and gravel runs, which are likelyformer active riffle locations that connect vegetated channel bars, shallow pools <2 m deep, and112.2. Field measurementslarge boulder cascade/riffle units. The lower reach is primarily composed of cobble and gravel withminimal fines and fairly uniform riffle/run morphology. The system has relatively low sedimentinput, except for two gravel augmentation piles and apparent debris flows in the upper reach. A100-m-long 0.6 ha plunge pool is located just downstream of the dam with a depth >2 m.Since dam construction in the 1920s, approximately 50 to 90% of historical Alouette River flowhas been diverted to the Stave River system for power generation (BCRP, 2000). The flow rateis currently maintained at approximately 3 m3/s throughout the summer months (Caryula andGroves, 2009). Discharge is primarily through a low level outlet (LLO) except between April 15and June 14, when a spring surface release of >3m3/s is discharged over the spillway to allowmigrating juvenile kokanee/sockeye to leave the lake and potentially re-anadromize (Caryula andGroves, 2009). The LLO is located at a depth between 6 to 10 m below the lake surface, which isoften within a metalimnetic layer during summer stratified periods. Strong winds frequently driveinternal seiching resulting in an approximately 12 h oscillation of warm and cold water at the LLO.Alouette River currently supports variably sized runs of all of the Pacific salmon species (sockeye,pink, chum, coho, chinook, and steelhead), as well as resident cutthroat, rainbow, and bull trout(FWCP, 2011). Stream temperature often exceeds salmonid thermal thresholds throughout thestudy reach (Greenbank, 2009), especially during the dry period of the summer when low flows anda warm epilimnetic reservoir release compound natural warming processes.2.2 Field measurementsData collection was conducted from early May to early October, 2013, to capture the period whentemperatures were previously observed to exceed salmonid thresholds. The first week consisted ofa site reconnaissance to identify features of interest and identify locations for instrument installa-tion and repeated manual measurements. Mainstem and tributary temperatures, conductivities,and above-stream microclimate variables were measured to develop an understanding of spatialheterogeneity. Instrument accuracy information is provided in Table 2.1.122.2. Field measurementsTable 2.1: Equipment used for data collection.Variables Sensor Range AccuracyWater temperature Onset Tidbit v2 -20 to 70 \u00E2\u0097\u00A6C 0.2 \u00E2\u0097\u00A6CAir temperature Rotronic HC-S3 -30 to +60 \u00E2\u0097\u00A6C 0.2 \u00E2\u0097\u00A6C(open site)Relative humidity Rotronic HC-S3 0 to 100% 1.5% @ 23 \u00E2\u0097\u00A6C(open site)Stream velocity Flo-Mate Model 2000 -0.15 m/s to +6 m/s 2%Shortwave radiation Kipp & Zonen CM6B 300 to 2800 nm <5%Electrical conductivity WTW TetraCon 325 1 \u00C2\u00B5S/cm to 500 mS/cm 1.5%Air temperature Kestrel 4500 -29 - 60 \u00E2\u0097\u00A6C 0.5 \u00E2\u0097\u00A6C(above-stream) Pocket Weather Tracker 0.5Relative humidity Kestrel 4500 5-95% 0.03%(above-stream) Pocket Weather TrackerWind speed Kestrel 4500 0.3-40 m/s 0.03 m/s(above-stream) Pocket Weather TrackerWind speed (open site) RM Young Wind Monitor 0-100 m/s 0.3 m/sBed temperature Omega HH-25TC -10 to 50 \u00E2\u0097\u00A6C 0.2 \u00E2\u0097\u00A6CNet radiation Kipp and Zonen NR 200 - 100 nm 5 to 10%Lite2 Net Radiometer2.2.1 Stream and bed temperatureMainstemMainstem temperature was continuously recorded at 10-min intervals at 25 locations T(1-25)(Figure 2.2). Onset Tidbit V2 temperature data loggers were fitted with radiation shields consistingof white PVC with holes drilled through to enhance water flow as recommended by Quilty and Moore(2007). Loggers were deployed in well mixed locations bracketing features that had the potentialto affect heat exchange rates. The largest six tributaries were bracketed, with one Tidbit 5-10 mupstream of the confluence, and one downstream where minor lateral temperature and conductivitygradients suggested reasonable mixing. Other features bracketed included a large pool at 1.6 km,a poorly vegetated gravel bank at 1.8 km, a large marsh at 2.7 km, the kame area, and a cutoffchannel from 9.5-10.3 km (Latimer Channel). Tidbits were also installed immediately downstreamof the LLO and plunge pool, near the confluence with Pitt River, and as required to ensure aspacing of <1 km along the study reach. The Tidbits were installed in early May and removed inlate September, except for five installed in July to correspond with bed Tidbits. A re-assessment oflateral temperature variation in early June resulted in relocation of four Tidbits by approximately10 m.132.2. Field measurementsFigure 2.2: Location of mainstem Tidbit temperature loggers (Google Earth TM, 2014).TributariesTidbit loggers were installed in the same fashion as the mainstem near the mouth of 11 trib-utaries for various durations over the course of the study period. Four of the tributaries weremeasured continuously throughout the study period to achieve a continuous time series, while theother seven tributaries were monitored for at least one month. Temperatures of unmonitored trib-utaries and upstream locations on monitored tributaries were measured sporadically with a WTWTetraCon 325 conductivity probe to gain a sense of spatial heterogeneity.Specific focus was placed on the kame area, where colder water and higher tributary flows wereobserved during preliminary assessment, which was expected to have a substantial cooling effecton Alouette River. The two tributaries with highest flow in this area were monitored continuously,142.2. Field measurementswhile five others were monitored for durations of at least one month. One seep emerging from asill in the hillslope was consistently within 0.5 \u00E2\u0097\u00A6C of 8 \u00E2\u0097\u00A6C, although these cool seeps appeared tobe few and quickly warmed as they flowed over the ground surface before reaching Alouette River.Bed temperatureBed temperature profiles were monitored using thermocouple wires mounted at depths of 5,20, and 40 cm on a wooden stake inserted into the stream bed at five subsurface monitoringsites (Figure 2.3). One additional profile was installed near the bank in the kame region wheregroundwater seepage was expected. Thermocouples were measured on up to seven separate dayswith a digital thermocouple reader (OmegaTMHH-25TC). One profile in the residential area wasapparently stolen, which we did not replace.Continuous time series of stream bed temperature were measured with buried Tidbits fromlate-July to late-September. Burial depths ranged from 29-36 cm at each of the five subsurfacemonitoring sites. Two of the bed Tidbits were in riffles and three were in shallow pool/runs.Recording intervals were 10 min.152.2. Field measurementsFigure 2.3: Map of subsurface monitoring sites (red) and geometric shade transects (green) (GoogleEarth TM, 2014).2.2.2 Streamflow and stream-subsurface exchangesReservoir discharge data were provided by BC Hydro for the full study period. These data includedthe proportion of flow released through the LLO and over the spillway. Flow data at the downstreamextent of the site were retrieved from the Water Survey of Canada (WSC) gauging station AlouetteRiver near Haney 08MH005, located at the 224th street crossing. These two datasets provided theboundary conditions for mainstem flow, which were augmented with measurements of discharge onthe mainstem at 11 locations and measurements of discharge at selected tributaries.162.2. Field measurementsMainstem flowVelocity-area measurements were collected with a Marsh-McBirney Flo-Mate Model 2000TMelectromagneticflow meter. The standard mid-section method (Lane, 1999) was applied to calculate discharge.Streamflow was calculated a total of 45 times on 14 separate days from May to September (Figure2.4). Measurements were repeated frequently at three sites with relatively uniform hydraulic con-ditions located at a distance downstream of the dam of 2.66 km (9 days), 5.68 km (11 days), and11.17 km (9 days).Figure 2.4: Location of gauged tributaries (blue) and mainstem flow measurement stations (red)(Google Earth TM, 2014).Turnipseed and Sauer (2010) suggested that standard errors between 3-6% can be expected172.2. Field measurementsgiven average measuring conditions. One streamflow measurement was made at the 224th StreetWSC gauging station, and was found to differ from the published WSC discharge by 0.14%. Thisgood agreement indicates that my manual measurements are consistent with WSC standards. Itshould be noted that the WSC gauging location had more uniform hydraulic geometry than theother sites, which may have allowed for a more accurate streamflow measurement.Tributary streamflowA total of 102 tributary streamflow measurements were taken over 15 days on 19 differentstreams (Figure 2.4). Streamflow measurements consisted of 53 salt dilution measurements, 17bucket collection measurements, and 32 stage measurements.Salt dilution measurements involved injecting salt solution at a constant rate and recordingthe steady-state electrical conductivity at a point downstream at which the solution was uniformlymixed across the stream channel, following the procedures outlined by Moore (2004). Suitabilityof sites for salt dilution measurements was established with qualitative observations of turbulenceand mixing and with measurements of lateral variations in conductivity and water temperature.Sites that did not reach a reasonable state of lateral homogeneity during trial runs were rejected.Bucket collection consisted of capturing culvert discharge with a graduated plastic 4 l containerand recording the time required for filling. This procedure was repeated at least three times andaveraged to attain a flow measurement.To facilitate frequent determination of tributary discharge in an efficient manner, stage-dischargerelations were established at nine tributaries to allow discharge to be estimated from stage obser-vations during brief site visits in between visits when discharge was measured. Stage was measuredrelative to the top of a piece of rebar that was installed vertically into the stream bed. Stage wasmeasured during each discharge measurement to construct the rating curve. An effort was made tomeasure tributary flow at different times of the day to smooth out the diurnal variation associatedwith evapotranspiration.Subsurface flowAn effort was made to characterize the subsurface exchanges including groundwater inflow,hyporheic downwelling and upwelling, and mainstem losses. Hyporheic exchange can be one ofthe more difficult flow paths to identify (Moore et al., 2005a; Carrivick et al., 2012), requiringextensive field work (Story et al., 2003). It was expected that, given the relatively uniform channelgradient and relatively low sinuosity (1.17), hyporheic exchange would be minimal (Moore et al.,2005a; Leach and Moore, 2011). Therefore, less attention was placed on characterizing hyporheicexchange than the other heat exchange processes.Piezometers were installed at each of the five subsurface monitoring sites to estimate hydraulicgradient in the bed (3 piezometers) and banks (9 piezometers) following the installation procedure182.2. Field measurementsoutlined by Baxter et al. (2003). Stream bed piezometers were only installed in locations withrelatively fine substrate to a depth of 21-42 cm. The head level and temperature at the base ofthe piezometer were measured up to six times at each site. Head level was measured in the bankand bed piezometers relative to adjacent stream water surface elevation. The head attributedto capillary action in the piezometers was assessed in the lab with a container of water and wassubtracted from the field measurements to correct for this effect.Patches of hyporheic downwelling are often characterized by weak temperature gradients (Silli-man and Booth, 1993) while groundwater discharge and hyporheic upwelling are often characterizedby a stronger temperature gradient with lower temperatures at depth (Moore et al., 2005a). It wastherefore anticipated that measuring profiles of bed temperature would also provide informationabout hyporheic exchange and groundwater upwelling. Spot temperature and conductivity mea-surements were also taken occasionally to identify in-channel groundwater upwelling sites.2.2.3 Hydraulic geometryHydraulic geometry was extracted from flow measurements and measurements taken at the shadetransect locations. Data relevant to the stream temperature model were measured including ve-locity, width, depth, and cross-sectional area. High resolution geometries were acquired from theflow transects, which included >20 depth and velocity measurements per transect at each of the11 sites. Three to eight depth measurements were also taken at the major slope breaks for each ofthe 63 shade transects. Cross-sectional area at these transects and streamflow measurements wereused to calculate mean velocity using the relationship v = Q/A.2.2.4 MeteorologyOpen site climateOne meteorological station was installed on top of the dam approximately 15 m above AlouetteRiver and 3 m above Alouette Lake to monitor open site meteorological variables. Sensors werescanned every second and averaged over 10-min intervals with a Campbell Scientific CR10X datalogger. Shortwave radiation was measured with a CMP3 Kipp and Zonen pyranometer. Temper-ature and relative humidity were measured with a Rotronic HC-S3 probe fitted with a radiationshield, and wind speed was measured with an RM Young Wind Monitor. Instrument accuraciesare provided in Table 2.1. Precipitation data and backup measurements of the aforementionedmeteorological variables were acquired from a meteorological station located at an open site in theMalcolm Knapp Research Forest approximately 5 km west of Alouette Dam at an elevation of 320m above mean sea level.192.2. Field measurementsAbove-stream MicroclimateTwo Kestrel 4500 Pocket Weather Trackers were deployed throughout the forested reach 1.5 mabove the stream surface to measure wind speed, relative humidity, and air temperature. One of theKestrels was meant to be left in the same place continuously but was stolen. The remaining Kestrelwas fastened to a PVC pipe and suspended over the channel for 1-3 week periods in nine differentlocations within the upper forested reach (Figure 2.5). Meteorological variables were scanned everysecond and averaged every 10-30 min. Kestrel specifications are provided in Table 2.1.Figure 2.5: Kestrel above-stream microclimate stations (Google Earth TM, 2014).Net radiationNet radiation was measured 35 cm above the stream surface using a Kipp and Zonen netradiometer. Two locations were measured for 8-10 h on Sept 3 and Sept 12, primarily to calibratethe hemispherical photo model. A Campbell Scientific CR10X logger was used to scan the netradiometer every second and average every minute. The net radiometer setup was not left outcontinuously due to risk of theft.202.2. Field measurements2.2.5 Riparian canopy characteristicsShade transectsRiparian vegetation characteristics were measured at 63 transects, including six islands, insupport of a geometric shade model. Transects were in clusters of two or three, spaced two channelwidths apart, at 24 locations spaced approximately 500 m apart, Figure 2.3. A summary of themeasurements taken is provided in Table 2.2. Vertical angles were measured with a clinometerwhile widths were measured with a meter stick or measuring tape.Vegetation characteristics were averaged over one channel width upstream and downstream.Tree heights and canopy overhang measurements were taken as the mid length between wherevegetation was uniformly opaque and the outer edge of foliage. Shade transects were measuredbetween August 7 and September 2, when leaves were anticipated to be at maximum size.Table 2.2: Shade transect measurements taken.Full site Widths for each bank (m) Angles for each bank (degrees)Wetted width waters edge to top of bank adjacent bank to top of bankPhotos in each direction waters edge to tree trunks adjacent bank to tree topsGeomorphic unit undergrowth overhang adjacent bank to undergrowth topVegetation class waters edge to edge of canopyVegetation species undercut lengthHemispherical photographsCalibration of the shade model was conducted using hemispherical photos analyzed with GapLight Analyzer software (Frazer et al., 1999; Leach and Moore, 2010). Photos were taken 30-40 cmabove the stream at 18 of the shade transects, with one photo in the centre and one approximately1.5 m from each bank. Photos were captured with a Nikon Coolpix 4500 4.0 mega pixel digitalcamera with a Nikon fisheye FC-E8 lens that was mounted on a leveled tripod and oriented north.Photos were collected between September 3-12, 2013, during periods of overcast skies or at dawnor dusk when sky conditions were uniform.21Chapter 3Analysis3.1 Time series analysis of stream temperatureThe following signals were expected to be found in the mainstem temperature time series: (1) sea-sonal variations, (2) synoptic-scale variations associated with weather systems, (3) diurnal signalsassociated with variations in solar radiation, (4) the effects of the downstream propagation of theinternal seiche in the reservoir, and (5) a tidal signal associated with tidally forced variations in PittRiver water levels. This section describes the analytical approaches used to separate each signalfrom the recorded temperature time series.3.1.1 Seiching signalFluctuations of reservoir discharge temperature in previous years have been attributed to an internalseiching process, whereby a wave was created in the interface (thermocline) between the warmsurface layer (epilimnion) and cool deep layer (hypolimnion) (Jackson, 1833). The seiche wascaused by wind shear acting over the lake, which pushes up water to the downwind side, resultingin a counteracting internal pressure gradient caused by the density difference of the epilimnion andhypolimnion (Fischer, 1979). The objective of the seiche analysis was to determine the strengthand period of the hypothesized internal seiche to see if the predicted temperature variations agreewith those observed at the outlet. Monthly profiles of lake temperature were provided by the BCMinistry of the Environment, with temperatures measured every metre from the surface to a depthof 20 m. BC Hydro had agreed to provide lake temperature profiles with 1-h resolution, but thisdata was lost due to logger theft.The period of the internal seiche, Ti (s), is dependent on stratification regime and lake geometry(Stevens and Lawrence, 1997):Ti =2L\u00E2\u0088\u009Ag\u00E2\u0080\u00B2h1h2(h1+h2)(3.1)where L is the length of the lake in the windward direction (m), h1 is the thickness of the epil-imnion (m), h2 is the average thickness of the hypolimnion (m), and g\u00E2\u0080\u00B2 is the reduced gravitationalacceleration (m/s2) based on density stratification:223.1. Time series analysis of stream temperatureg\u00E2\u0080\u00B2 = g(\u00CF\u00812 \u00E2\u0088\u0092 \u00CF\u00811)/\u00CF\u00812 (3.2)where \u00CF\u00811 is the density (kg/m3) of the epilimnion and \u00CF\u00812 is the density (kg/m3) of the hypolimnion.Densities were calculated based on July temperatures and the temperature-density relation de-scribed by Vallentyne (1957):\u00CF\u0081(T ) = 1000\u00E2\u0088\u0092(T \u00E2\u0088\u0092 4)2150(3.3)The deflection at the outlet, \u00CE\u00B6i (m), is dependent on wind speed, stratification and reservoir geom-etry as follows (Stevens and Lawrence, 1997):\u00CE\u00B6i =u2\u00E2\u0088\u0097L2g\u00E2\u0080\u00B2h1(3.4)where u\u00E2\u0088\u0097 is the shear velocity (m/s):u\u00E2\u0088\u0097 =\u00E2\u0088\u009ACD\u00CF\u0081aU210\u00CF\u0081o(3.5)where CD is a drag coefficient set to 10\u00E2\u0088\u00923 (Stevens and Lawrence, 1997), \u00CF\u0081a is air density (kg/m3),\u00CF\u0081o is a reference density set to 1000 kg/m3 (Fischer, 1979), and U10 is wind speed (m/s) acting 10m above the lake surface. Spigel and Imberger (1980) found that wind needs to be acting for aduration of Ti/4 to set up a seiche. Therefore, U10 was calculated as the average of the daily meansof highest wind speed acting for Ti/4 consecutive hours.3.1.2 Wavelet analysisWavelet analysis can be used to represent a signal in time-frequency space, which shows the strengthof signals with varying frequencies over different time periods (Torrence and Compo, 1998). Apreliminary wavelet transform was performed on the reservoir outlet temperature time series (T1)to assess the strength and duration of the seiching signal in order to guide further investigation.This was followed by a more detailed analysis including wavelet transforms on temperature timeseries from locations T2, T7, T12, T15, T22, T23, and T26 and cross-wavelet analysis of these timeseries with T1 to determine if the seiching signal at T1 could be detected at downstream locations.The Morlet wavelet function was applied due to its frequent use in analyzing geophysical data(Torrence and Compo, 1998; Lafreniere and Sharp, 2003; Salmond, 2005; Liu et al., 2007; Veledaet al., 2012); it has the following form:\u00CE\u00A8jn = 2\u00E2\u0088\u0092j/2\u00CE\u00A8(2jt\u00E2\u0088\u0092 n) (3.6)where j represents the number of translations, n is the number of scaling indices, and \u00CE\u00A8 is thewavelet function (Veleda et al., 2012). The quantities n and j are related to time by the equation233.1. Time series analysis of stream temperaturet = n2j . A continuous wavelet transform was then applied to each temperature time series T (t) asfollows:f(u, s) =+\u00E2\u0088\u009E\u00E2\u0088\u00AB\u00E2\u0088\u0092\u00E2\u0088\u009ET (t)\u00CE\u00A8\u00E2\u0088\u0097u,sdt (3.7)where \u00CE\u00A8\u00E2\u0088\u0097 is the complex conjugate of \u00CE\u00A8u,s, s = 2j , and u = n2j . The wavelet transform canbe defined as the convolution of T (t) with a translated and scaled version of \u00CE\u00A8u,s (Torrence andCompo, 1998). The wavelet transform must then be discretized to determine the contribution ofscale 2j at location n2j to T (t):f jn =+\u00E2\u0088\u009E\u00E2\u0088\u00AB\u00E2\u0088\u0092\u00E2\u0088\u009ET (t)\u00CE\u00A8jndt (3.8)The strength of the signal corresponding to each coefficient f jn can then be represented by the powerspectra:Enj = |2j/2f jn|2 (3.9)The inclusion of the scaling factor 2j/2 as a multiplier was a later development proposed by Liuet al. (2007), which removes the bias of the original function described by Torrence and Compo(1998). Previously, the transform coefficients were not divided by scale, resulting in inaccuratelyhigh power for longer periods (Liu et al., 2007). Cross-wavelet transforms were performed followingthe methods of Veleda et al. (2012), who proposed a similar method to remove low frequency bias:W j,kn = (2j/2f jn)(2k/2gkn)\u00E2\u0088\u0097 (3.10)where k and gkn correspond to the translation index (j) and the coefficient (fjn) of the secondwavelet transform. The asterisk(\u00E2\u0088\u0097) refers to complex conjugation (Veleda et al., 2012). The valueof W j,kn represents the degree of correlation of the signals corresponding to scale 2j at locationn2j (Veleda et al., 2012). Significance levels and confidence intervals were calculated with themethods proposed by Torrence and Compo (1998) for both wavelet transform power spectra andcross wavelet values. The significance test consisted of comparing the temperature time seriespower spectra to theoretical wavelet spectra for white-noise and red-noise processes (Torrence andCompo, 1998; Gouhier, 2014).3.1.3 Cross-correlation analysisAssuming steady flow, the seiching signal should propagate downstream at the mean velocity ofwater flow. The maximum cross-correlation between temperature signals should occur at a lag equal243.2. Stream temperature modelto the travel time of a water parcel. Cross-correlation involves computing the Pearson product-moment correlation coefficient between temperature time series at two locations with progressivelags for one of the locations. The R ccf{stats} function was applied to calculate cross-correlationfunctions (R Core Team, 2013).Cross-correlation functions were computed between T1 and the same downstream locationsassessed with the cross-wavelet analysis. Lags for every 10-min increment were compared up to atotal of 15 h for T2-T15, and 30 h for T22-T26. The lag intervals assessed were based on traveltime estimates based on measured velocities. The tidal signal was also assessed by calculatingthe cross-correlation function between the water surface elevation in New Westminster, which wasretrieved from the Department of Fisheries and Oceans Canada, and each of T22, T23, T26, andNorth Alouette River. This analysis provides an indication of how far upstream the tidal signalpropagated, with an upper bound of T22, where the tidal signal was assumed to be absent due tothe stream bed elevation.Prior to computing the cross-correlation function, each temperature time series was band-passfiltered for oscillation periods between 9-15 h. This ensured that temperature fluctuations associ-ated with the semi-diurnal seiching signal would be maintained, while the diurnal and most othersignals would be removed. A Christiano-Fitzgerald band pass filter was applied, which generates afiltered time series with the unrestricted optimal filter (Christiano and Fitzgerald, 2003; Balcilar,2007). A drift was allowed to account for the seasonal moving average .The time lag with maximum cross-correlation with T1 was determined for each Tidbit locationdownstream. This lag time provided an estimated flow travel time for the semi-diurnal signal topropagate downstream, which was used to determine time- and space- averaged velocity. Thesevelocities were compared with those estimated from the hydraulic geometry transects to generatea steady-state velocity estimate for the stream temperature model.3.2 Stream temperature modelThis section describes the structure and equations incorporated into the stream temperature model.It also describes how the meteorological and hydrological boundary conditions were specified.3.2.1 Model structureA physically based deterministic model was developed and applied to model Alouette River temper-ature. The model was based on the model described by Leach and Moore (2010, 2011). The modelis based on a Lagrangian reference frame, where a parcel of water (indexed by j) is released ateach 10-min interval (t) and is then tracked as it flows through a series of i nodes with spacing \u00E2\u0088\u0086x(m), corresponding to 10-min travel times. Between each node at each time step, the advective and253.2. Stream temperature modelnon-advective energy exchange rates are calculated, which are used to compute the downstreamtemperature of parcel j at node i+ 1 and time t+ 1 as follows:dTwdx=HWCQT+qr[Ttr(t)\u00E2\u0088\u0092 Tw(i)] + qh[Tbed(t)\u00E2\u0088\u0092 Tw(i)] + qg[Tbed(t)\u00E2\u0088\u0092 Tw(i)]QT(3.11)Tw(i+1,t+1) = Tw(i,t) + \u00E2\u0088\u0086xdTwdx(3.12)where W is the channel width (m), H is non-advective heat exchange (W/m2) at the stream surfaceand bed, C is the heat capacity of water (4.18\u00C2\u00B7106 Jm-3\u00E2\u0097\u00A6C-1), QT is total flow (m3/s), qr is tributaryinflow (m2/s), Ttr is tributary temperature (\u00E2\u0097\u00A6C), qg is groundwater inflow (m2/s), qh is hyporheicinflow (m2/s), and Tbed is bed temperature (\u00E2\u0097\u00A6C). Non-advective heat exchange H was calculatedas:H = K\u00E2\u0088\u0097 + L\u00E2\u0088\u0097 +Qh +Qe +Qc (3.13)where each term is in W/m2, K\u00E2\u0088\u0097 is net shortwave radiation, L\u00E2\u0088\u0097 is net longwave radiation, Qh issensible heat exchange, Qe is latent heat exchange, and Qc is bed heat conduction. The parcels weretracked from the plunge pool to the WSC gauge 14 km downstream for every 10-min incrementof the study period. The travel time was based on the steady velocity estimate of 0.35 m/s. Themodel was evaluated by assessing prediction accuracy at the downstream extent and at 6.8 km,where the channel gradient and riparian vegetation change and partly residential land use begins.Shortwave radiation and bed heat conduction were treated as spatially uniform within eight thermalreaches based on streamflow and shade conditions. The turbulent fluxes, Qe and Qh, and longwaveradiation were calculated as a function of the temperature of the water parcel being tracked andother study reach or thermal reach averaged variables described below. The relative contributionof each heat flux was summed for each parcel of water from the time it left the plunge pool to itsarrival at the WSC gauge. Each heat flux was multiplied by the time step for each node to findthe total energy gained or lost:Ej,f =t=n\u00E2\u0088\u0091t=1\u00E2\u0088\u0086tQf,t (3.14)where Ej,f is the net heat absorbed (joules) by parcel j, \u00E2\u0088\u0086t is the travel time between nodes (s),and Qf,t is the heat flux (W/m2) from process f at time t.3.2.2 Above-stream microclimateCalculation of the surface energy exchanges requires values for air temperature (Ta), vapour pressure(ea), and wind speed (Uw) above the stream surface. The roving Kestrels provided values atnine different sites within the upper forested reach; however, when the Kestrel was exposed to263.2. Stream temperature modeldirect sunlight it over-heated and recorded unrealistically high Ta values. Therefore, only datafrom three fully shaded locations accounting for seven weeks were considered accurate. Above-stream microclimate variables for the remainder of the period were predicted using linear regressionswith variables measured at the open site meteorological station. The relations between valuesmeasured at the Kestrel sites with the open site were similar, so data from all three sites werepooled to generate a spatially averaged regression model for each variable. The regression modelswere determined by finding the best fit based on different combinations of the available variablesand transformations.Only one Kestrel wind speed time series was compromised due to spiders jamming the anemome-ter. The lost period was filled in with a linear regression using open site wind speed and Ta aspredictor variables. The Kestrel stall speed of 0.3 m/s was accounted for by adjusting zero valuesto half the stall speed, and to zero when the dam wind speed was less than 0.1 m/s. A time serieswith 10-min increments was prepared by linearly interpolating between measurements when therecording interval was >10 min.3.2.3 Turbulent exchangesLatent heat exchange was calculated using the following Dalton-type equation:Qe = 28.36(2.64 + 2.92Uw)(ea \u00E2\u0088\u0092 ew) (3.15)where Uw is the wind speed (m/s) 1.5 m above the water surface, ea is atmospheric vapour pressure(kPa) and ew is the vapour pressure (kPa) at the water surface.The coefficients in Eq. (3.15) were taken from those derived for Catamaran Brook, NB, Canada(Maheu et al., 2013), because that stream was closest in width to Alouette River than other streamsat which wind functions have been derived, as summarized by Guenther et al. (2012). Atmosphericvapour pressure, ea, was calculated following the methods of Stull (2000):ea =(RH100)esat(Ta) (3.16)where esat is the saturation vapour pressure (kPa):esat(T ) = eo \u00C2\u00B7 exp[Lv 0.80(3.22)where kt is the ratio of incident solar radiation to solar radiation at the top of the atmosphere,which is a function of solar constant and solar zenith angle (Erbs et al., 1982).Net longwave radiation was calculated as:L\u00E2\u0088\u0097 = \u000FwL \u00E2\u0086\u0093 \u00E2\u0088\u0092L \u00E2\u0086\u0091 (3.23)where L\u00E2\u0086\u0093 is the longwave radiation received at the stream surface, L\u00E2\u0086\u0091 is the longwave radiationemitted by the water surface, and \u000Fw is the emissivity of the water, assumed to equal 0.95 (Mooreet al., 2005b). Longwave radiation received at the stream surface was calculated as:L \u00E2\u0086\u0093= [fv\u00CE\u00B5a + (1\u00E2\u0088\u0092 fv)\u00CE\u00B5vt]\u00CF\u0083(Ta + 273.2)4 (3.24)283.2. Stream temperature modelwhere \u00CE\u00B5a is the atmospheric emissivity, \u00CE\u00B5vt is the emissivity of vegetation and terrain, assumed toequal 0.97 (Moore et al., 2005b), and \u00CF\u0083 is the Stefan-Boltzmann constant (5.67 \u00C2\u00B7 10\u00E2\u0088\u00928Wm\u00E2\u0088\u00922K\u00E2\u0088\u00924).A spatially averaged atmospheric emissivity time series was calculated as a function of atmosphericwater content and cloud cover using the methods of Leach and Moore (2010). Emitted longwavefrom the stream surface was calculated as:L \u00E2\u0086\u0091= \u00CE\u00B5w\u00CF\u0083(Tw + 273.2)4 (3.25)Direct shortwave modelA geometric model for computing direct shortwave radiation was adapted from the SHADEmodel developed by Chen et al. (1998a,b), which accounts for solar position, stream orientation,channel width, bank vegetation height, width from bank to tree tops, and overhang length forriparian vegetation. Stream orientation was calculated using stream centerline northing and eastingcoordinates at each grid cell of a 25 m resolution raster digital elevation model (DEM). Overhanglength was assumed to be the maximum of either the overhanging underbrush or canopy. Bankheights were not modelled due to their relatively low height and lack of effective shade.The model calculates the shaded fraction of the channel as a function of time. The total shadelength cast by the riparian vegetation, s, parallel to the solar azimuth, was modelled as follows:s = HTss/ tan(pi2\u00E2\u0088\u0092 \u00CE\u00B8) (3.26)where HTss is the tree height on the sun side of the channel (m) and \u00CE\u00B8 is the solar zenith angle(radians). The sun side of the channel is either left or right corresponding to the side nearest tothe sun, or center when the solar azimuth angle, \u00CF\u0088, is equal to the channel orientation angle \u00C2\u00B1 theangle between vertical and the top of vegetation on either bank from the channel center line. Thesolar angles, \u00CE\u00B8 and \u00CF\u0088, were calculated using the equations described in Iqbal (1983). The shadedfraction of the channel width, sp, was then modelled as:sp = s cos(pi\u00CF\u0089180)\u00E2\u0088\u0092WTss (3.27)where \u00CF\u0089 is the acute angle (degrees) between a line perpendicular to the stream and the azimuth,and WTss is the width to tree tops on the sun side of the channel (m). The geometrically modelleddirect shortwave function g(t) was calculated as a function of perpendicular shade fraction, sp,vertical vegetation transparency Ttr, overhang width Ow, overhang transparency Otr, and channelwidth W (m) as follows:g(t) = 1\u00E2\u0088\u0092(Ttr \u00C2\u00B7 sp+Otr \u00C2\u00B7Ow,ssW)(3.28)293.2. Stream temperature modelThe centered sun side condition when \u00CE\u00A8 was aligned with the channel orientation was modelledseparately as follows:g(t) = 1\u00E2\u0088\u0092(Otr(Ow,R +Ow,L)W)(3.29)The overhang transparency, Otr, was modelled as:Otr =Kdir,op(t)\u00E2\u0088\u0092Kdir,hemis(t)Kdir,op(t)(3.30)whereKdir,op(t) is the direct shortwave radiation (W/m2) measured at the open site andKdir,hemis(t)is direct shortwave radiation modelled with the hemispherical photographs described below. Mod-elling of the gap fraction was an iterative process to identify the vegetation transparencies Ttr andOtr. Bank vegetation was similar on both sides of the channel, so transparencies were averaged foreach reach regardless of channel side. Transparency was calculated independently in the residentialareas with partial bank tree removal due to relatively thin riparian buffer width. The value of g(t)was assumed to equal zero when the solar elevation angle (1\u00E2\u0088\u0092\u00CE\u00A8) was less than the horizon angle,which is the angle between horizontal and the top of surrounding terrain. The horizon angle wascalculated as described in Richards et al. (2012) using the 25 m resolution DEM. An additional 32m of elevation was added to the valley walls and the surrounding area to simulate the approximately80 year old conifer-dominated forest that surround the area (GVRD, 2004).View factor modelThe view factor model presented in Moore et al. (2014) was adjusted to account for the localconditions of each shade transect. The model calculates the width-averaged view factor as a func-tion of bank height, bank vegetation height and underside height, channel width, and overhanglength (Moore et al., 2014). A variation of Case 4 of Moore et al. (2014) was applied when over-hanging vegetation was present on both banks, while Case 5 was applied when one bank had offsetor absent riparian vegetation and the other had overhanging vegetation.Hemispherical photograph modelThe geometric shortwave and sky view factor models were calibrated by comparing their pre-dictions to those made using hemispherical photographs and the model described by Moore et al.(2005b) and Leach and Moore (2010). Formally, the sky view factor can be expressed as:fv =1pi\u00E2\u0088\u00AB 2pi0\u00E2\u0088\u00AB pi/20g\u00E2\u0088\u0097(\u00CE\u00B8, \u00CF\u0088) cos \u00CE\u00B8 sin \u00CE\u00B8 \u00C2\u00B7 d\u00CE\u00B8 \u00C2\u00B7 d\u00CF\u0088 (3.31)where g\u00E2\u0088\u0097(\u00CE\u00B8, \u00CF\u0088) is the gap fraction at sky position (\u00CE\u00B8, \u00CF\u0088). In practice, the integral was calculatedfrom gap fractions computed for a 5\u00E2\u0097\u00A6 x 5\u00E2\u0097\u00A6 grid of zenith (\u00CE\u00B8, vertical = 0) and azimuth (\u00CF\u0088) angles303.2. Stream temperature modelusing the Gap Light Analyzer software (Frazer et al., 1999). The double integral was approximatedas the summation of the function for each grid cell.Sky view factors were calculated under the right canopy, center, and left canopy locations ateach shade transect using the hemispherical photographs. These values were averaged for each tran-sect based on the relative channel width occupied by those regions. The geometrically modelledg(t) function was compared for each transect with modelled values from the hemispherical pho-tographs for a cloudless day. The accuracy of the view factors and g(t) functions derived from thehemispherical photographs was assessed by comparing modelled net radiation with measurementsmade at the same locations as two of the hemispherical photos.Shade reachesShade reaches were delineated based on changes in the direction of the central tendency of thestream and the beginning of the residential area. Five reaches were identified, with two in theforested section with a roughly southwestern stream orientation, and three in the residential areawith a roughly western orientation. The averages of shade characteristics along de-forested banksand fully vegetated banks were calculated separately. The fractions of each shade reach with de-forested banks were measured using Google EarthTMfor each residential reach, which was used tocalculate a weighted average for each shade characteristic. Some stretches of the stream featureda split channel for lengths of 50 - 250 m with heavily vegetated islands separating them. In theselocations, the shade characteristics were modelled for both channels, then combined into one witha width-weighted average.3.2.5 Tributary flowTributary subcatchments were delineated for each tributary joining between the dam and WSCgauge (Figure 3.1) using SAGA GIS (Cimmery, 2010). A multiple flow direction method (i.e. flowdistribution with divergence) called DEMON was applied to generate the channel networks. Thedrainage area required to initiate a channel was set to 100,000 m2, which agreed well with tributaryGPS points from the field. The smallest allowable channel length was set to 6 cells or 150 m.313.2. Stream temperature modelFigure 3.1: Tributary subcatchments delineated with SAGA GIS.For each of the creeks with rebar staff gauges, a regression was performed to develop a stage-discharge rating curve. If a stage measurement showed either an increase in flow and increasedemergent rebar length, or a decreased flow and decreased emergent rebar length, it was removed.This occurred four times at two locations where a nearby culvert or channel constriction becamepartially jammed with debris. The gauged tributaries were identified on the subcatchment map(Figure 3.1) and the drainage areas were calculated at the specific locations of flow measurements.Each flow measurement was divided by drainage area to compute a unit flow (Q/DA). To generatea continuous time series of tributary inflows, each unit flow was regressed against the correspond-ing streamflow at the WSC gauge. The tributaries exhibited a range of Q/DA regimes leading tothe identification of regions of similar runoff regime, highlighted in different colours on Figure 3.1.Cumulative tributary inflow was calculated at the three repeated mainstem flow measurement loca-tions at 10-min intervals for the full study period. Tributary inflows were treated as an equivalentaverage input per unit channel length for each discharge reach, rather than as discrete inputs. This323.2. Stream temperature modelapproximation was assumed to be reasonable for the purpose of the stream temperature modelsince no tributary made up more than 0.4% of total inflow.3.2.6 Mainstem flow calculationThe fraction of the full study reach inflow that had come in above the location of each flow meteringstation along the study length (RQstn) was calculated as follows:RQstn =Qstn \u00E2\u0088\u0092QdamQWSC \u00E2\u0088\u0092Qdam(3.32)where Qstn is the measured flow at each metering station, Qdam is the dam discharge, and QWSCis the flow at the WSC gauge. Assuming continuous inflows and no losses from the system, RQstnshould be a factor between 0 and 1. The factor was assumed temporally constant at each stationdue to the lack of high flow measurements and relatively steady flow over the summer.Values of RQstn were only calculated for stations with >5 flow measurements due to the rel-atively high measurement uncertainty at the gauging locations. During the period of spillwaydischarge and dry mid-August weather, RQstn was occasionally <0 or >>1, suggesting flow losses.The observed surface runoff and relatively low hyporheic exchange potential of the channel indi-cated this was improbable and likely due to errors in Qdam. Therefore, RQstn values outside therange 0-1 were removed from the averaging.The raw QWSC data exhibited sub-hourly noise with variations between 0.1 - 0.2 m3/s. Thisnoise was assumed to be measurement error related to turbulence or surface waves affecting theWSC pressure transducer. To remove this noise while maintaining the diurnal evapotranspirationsignal, a loess smoothing function was applied with a span of 0.1 (R Core Team, 2013).3.2.7 Groundwater distributionGroundwater was assumed to equal the difference between total inflow (Qstn\u00E2\u0088\u0092Qdam) and cumulativetributary inflow between each pair of mainstem flow stations. This value was often negative,especially in the dry period of the summer, indicating a loss of flow. It was initially intendedto partition the total inflow at each station into either surface runoff or groundwater; however,the uncertainty of RQstn values was estimated to be higher than tributary runoff uncertainty.Therefore, the modelled cumulative tributary inflow was assumed accurate and the channel wasassumed to be losing flow by infiltration into the bed or banks when Qstn \u00E2\u0088\u0092Qdam was negative.3.2.8 Hyporheic exchangeDarcy\u00E2\u0080\u0099s law was applied to estimate hyporheic exchange following Moore et al. (2005b), as follows:qz = Ksat\u00E2\u0088\u0086h/\u00E2\u0088\u0086z (3.33)333.2. Stream temperature modelwhere (Ksat) is the saturated hydraulic conductivity (m/s), estimated from relations with streambed composition (Freeze and Cherry, 1979), and \u00E2\u0088\u0086h/\u00E2\u0088\u0086z is the vertical hydraulic gradient deter-mined from piezometer measurements (m/m).Hyporheic exchange was considered active during periods when downwelling was observed andthe stream was found to be losing. When the stream was gaining from groundwater discharge, bankto stream and stream to bed hydraulic gradients were positive, which would suppress hyporheicexchange. Stream bed temperature measurements from the buried Tidbits and bed temperatureprofiles also provided evidence of hyporheic exchange. Zones of hyporheic downwelling are oftencharacterized by greater bed temperature diurnal variation, and weak temperature gradients (Silli-man and Booth, 1993; Moore et al., 2005b; Leach and Moore, 2011), while groundwater dischargeand hyporheic upwelling are often characterized by a stronger temperature gradient with coldertemperatures at depth (Moore et al., 2005a).3.2.9 Inflow temperatureTributary inflow temperature was estimated as the average over all measured tributaries. Thehyporheic and groundwater temperatures were taken as the average of the bed Tidbits within eachof the four flow reaches separated by the flow measurement stations. Bed Tidbits were installed inAugust so bed temperatures for the earlier months were estimated via linear regression with streamtemperature at the nearest stream Tidbit. The bed temperatures were not expected to have a stronglinear relation to stream temperature during different hydrological periods; however, the regressionis assumed conservative since it is based on the period of warmest stream temperatures.3.2.10 Bed heat conductionBed heat conduction (Qc) was calculated using the five stream bed Tidbits and overlying measuredstream temperatures by solving Fourier\u00E2\u0080\u0099s law of heat conduction. The stream and bed temperatureswere used as upper and lower boundary conditions, respectively, and the initial temperature distri-bution with depth was assumed to vary linearly from the upper to lower boundary. The equationswere solved using a forward-time/centered-space finite difference approach (Chapra, 1997). Tem-peratures were computed at nodes with 5 cm spacing (indexed by j) at 10-min intervals (indexedby i) as follows:Ti+1,j = Tij +E\u00E2\u0088\u0086t\u00E2\u0088\u0086x2(Ti,j+1 \u00E2\u0088\u0092 2Ti,j + Ti,j\u00E2\u0088\u00921) (3.34)where E is the thermal diffusivity, computed asE =Kc\u00CF\u0081bCp(3.35)343.2. Stream temperature modelwhere \u00CF\u0081b is the density of saturated sediments (kg/m3), and Cp is the specific heat of saturatedsediments (1130 Jkg-1K-1). Thermal conductivity was estimated to be between 1.8 to 2.8 Wm-1K-1based on an estimated porosity of 0.2-0.4 and sediment density of 2.58 g/cm3 (Freeze and Cherry,1979; Lapham, 1989). The mean of these values (2.3 Wm-1K-1) was applied in the model, with theupper and lower estimates used as uncertainty limits.Bed heat conduction was computed using the simulated temperature gradient between the bedsurface and the next lower node. These computed rates were averaged for each of the four flowreaches. Note that this approach is based on the assumption that advective bed heat transport byhyporheic exchange and groundwater discharge is negligible. Bed temperature profiles measuredat the thermocouple stakes were used as an informal test of modelled values.3.2.11 Uncertainty/sensitivity analysisMany of the required inputs to the model are subject to uncertainty. The overall impact of theseuncertainties on predicted stream temperatures was assessed by running the model not only forthe best available estimate of each input, but also for lower and upper limits, as listed in Table3.1. The medium value represents the best estimate of the input that was used in the mitigationscenarios.353.2.StreamtemperaturemodelTable 3.1: Summary of inputs used in model uncertainty analysis runs. The middle estimate represents the best estimate of the inputvariable or parameter. The low and high estimates represent the bounds of possible values for each input.Variable Affected heat flux Low Medium Highv (m/s) All 0.32 (est. from MPC) 0.35 (mean estimate) 0.38 (est. from transects)Kdir (factor) Kdir vegetation opacity of: vegetation opacity of: vegetation opacity of:1.0 (vert.), and 1.25 (horiz.) 0.8 (vert.), and 1.0 (horiz.) 0.6 (vert.), and 0.75 (horiz.)fv Kdiff 1.25 x fv 1.0 x fv 0.75 x fvfv, L \u00E2\u0086\u0093 (equation) L \u00E2\u0086\u0093 1.25 x fv, 1.0 x fv 0.75 x fv,(Chapra, 1997) lower coeff. (Moore et al., 2005b) (Chapra, 1997) higher coeff.Wind function equation Qe and Qh Griffith Creek Catamaran Brook Miramichi River(Guenther et al., 2012) (Maheu et al., 2013) (Maheu et al., 2013)Wind speed (factor) Qe and Qh 0.5 1 1.5Thermal conductivity Qc 1.8 2.3 2.8Kc (W/m \u00C2\u00B7K)Tbed Qc, advective exchanges coolest measured mean per flow reach warmest measuredTtr advective exchanges coolest measured mean per flow reach warmest measuredHydraulic conductivity qh 1E-5 1E-4 1E-3K (m/s)qr, qg, QT advective exchanges modelled without dam, based on RQ linear inflows alongqr based on tributary gauging, entire study reachqg = 0.3 x qrWidth (factor) surface exchanges 0.9 1.0 1.1363.3. Management scenariosThe resulting uncertainty in the predicted temperatures should be conservative because someinputs were varied by an unrealistic amount to test the model sensitivity to each value. For example,the upper and lower estimates of L \u00E2\u0086\u0093 were derived from the equation presented in Chapra (1997),which was intended for lakes. Also, the range of inflow and bed temperatures modelled assumeall tributaries and bed temperatures were either as warm as the warmest measured, or as cold asthe coldest measured. The variation of the advective exchanges represents three cases that couldrepresent reality, and do not necessarily indicate higher or lower heat exchange. The mediumcase represents the flow distribution, RQ, described in the previous section, where the total inflowwas partitioned into groundwater or tributary runoff. The low case neglects the flow data at thedam, since the accuracy was questionable, and assumes the streamflow at each metering stationwas the difference between the WSC gauge and the cumulative tributary inflow modelled betweenthe WSC gauge and the metering station. An additional groundwater input equal to 30% of thetributary inflow was assumed and similarly deducted from the WSC gauge streamflow. The highcase assumes that inflows were spatially uniform along the study reach, and the total study reachinflow predicted from RQ was linearly distributed.3.3 Management scenariosThree different management options were investigated through application of the stream tempera-ture model: (1) increased forest along channel banks, (2) lowered reservoir withdrawal level/depth,and (3) increased reservoir discharge. Details of each scenario follow.Bank re-forestationA walk of the site and aerial photograph analysis indicated that 9% of the residential areabanks had gaps in riparian trees that could be replanted, which corresponds to 4% of the totalstudy reach. Potential reductions in solar radiation were determined by assuming all banks wouldhave similar vegetation characteristics to adjacent fully forested areas within the residential area(shade reaches 3-5).Reservoir withdrawal levelThe relatively shallow intake of the LLO allows discharge temperatures to respond to solarenergy and diurnal oscillations of the boundary between the epilimnion and hypolimnion. Loweringof withdrawal depth to 15 m and 20 m was modelled to determine the effect of releasing coolerhypolimnetic temperatures. An epilimnetic release of 5 m depth was also modelled to simulate aworst-case scenario. Depths reasonably far from the metalimnion were chosen to avoid modellingof the seiching signal, which is dependent on wind conditions and could be difficult to accuratelycharacterize. Temperature profiles were interpolated between monthly measurements to obtain373.3. Management scenarios10-min resolution in lieu of a higher resolution time series.Discharge augmentation by increased flow releasesDischarge augmentation was assessed by adding 1, 2, 4, and 8 (m3/s) to the measured releaserate over the full study period. Modelling the effects of increased discharge required determiningassociated changes in hydraulic geometry including velocity, width, depth, and flow area. Velocitywas averaged over the entire study reach and study period, while width was averaged within each ofthe five shade reaches. Depth and flow area were inferred from relations to velocity, width, and flowrate. Substrate size (D84) was estimated to be 0.6 m in the steeper reach and 0.3 m in the lowerreach based on field observations. An average bank angle of 45o was assumed based on transectfield measurements. Average bank height was found to be 0.4 m above the typical water surfacelevel, which provided an upper bound for potential depth increase before residential properties maybe threatened by flooding.The effect of increased flow on channel width, depth, and velocity was determined as a functionof relative bed roughness following Ferguson (2007). Channels with readily erodible banks can alsowiden with increased baseflow, necessitating the use of regime models that consider bank stabilityindices (Eaton, 2006). This approach was assumed unnecessary since Alouette River\u00E2\u0080\u0099s banks aremainly composed of large cobbles remnant from pre-dam conditions when flows were an order ofmagnitude higher. Instead, the variable-power equation proposed by Ferguson (2007) was applied,which predicts bed friction with a power relation of depth and grain size at high flows, and a moreroughness-layer dependent relation at shallower flows:(8f)0.5= a1a2(dD)/[a21 + a22(dD)(5/3)]0.5 (3.36)where f is the Darcy-Weisbach friction factor, a1 and a2 are constants related to deep flows andshallow flows respectively, d is channel depth (m), and D is grain size (m) (D84). Velocity, v (m/s),can then be determined with the relation:vu\u00E2\u0088\u0097=(8f)0.5(3.37)where u\u00E2\u0088\u0097 is shear velocity (m/s):u\u00E2\u0088\u0097 = (gdS)0.5 (3.38)where g is acceleration due to gravity (9.8 m/s2) and S is channel gradient. Using the Fergu-son (2007) relation, an iterative process was applied in which a range of flows corresponding toincremental increases in depth were calculated until the flow corresponding to an increased damrelease scenario was achieved. The mean value of a1 was set to 7.5, which corresponds to the meanof values cited in the literature (7 to 8) (Ferguson, 2007). The a2 value was found numerically383.3. Management scenariosby back-calculating with observed flows at each of the cross sections and taking the mean. Allhydraulic geometry variables were calculated at each transect and averaged over the entire studyreach, except width, which was averaged for each of the five shade reaches.39Chapter 4ResultsThis chapter presents an overview of the meteorological and hydrological conditions of the studyperiod (Section 4.1), followed by an analysis of the observed stream temperature patterns (Section4.2). Measurements of the variables affecting stream temperature are then presented (Section 4.3),followed by a description of the data processing required to prepare heat flux time series for thestream temperature model (Section 4.4). An evaluation of the stream temperature model is thenpresented (Section 4.5), followed by a summary of the cumulative heat absorbed by parcels as theyflow through the study reach (Section 4.6). Lastly, an evaluation of the management scenarios ispresented (Section 4.7).4.1 Overview of the study periodThe 2013 study period was drier and warmer than the average of the previous 15 years (Table4.1). July temperatures were 0.9 \u00E2\u0097\u00A6C higher than average, and almost 1 \u00E2\u0097\u00A6C higher than each of theprevious three years. August temperatures were closer to the average temperature, and lower thanthe preceding year. Total precipitation was only 20% of the average for July and August, but was5% higher than the June to September average.Table 4.1: Historical air temperature and precipitation recorded at a BC Hydro managed weatherstation on Alouette Dam. Precipitation values for 2013 were not measured at the dam and in-stead were predicted from precipitation at Malcolm Knapp Research Forest (MKRF) with a linearregression.variable 1998 - 2012 2010 2011 2012 2013May mean Ta \u00E2\u0097\u00A6C 11.5 10.7 10.1 11.3 NAJune mean Ta \u00E2\u0097\u00A6C 14.5 13.5 13.9 12.6 14.8July mean Ta \u00E2\u0097\u00A6C 17.6 17.7 15.9 17.5 18.5August mean Ta \u00E2\u0097\u00A6C 17.8 17.9 17.8 19 18.1September mean Ta \u00E2\u0097\u00A6C 14.9 15.2 16.6 16.2 NAJun to Sep total precip (mm) 346 426 378 293 364Jul to Aug total precip (mm) 133 76 156 88 28Time series of meteorological variables measured during the study period at the dam, and404.1. Overview of the study periodMalcolm Knapp Research Forest in the case of precipitation, are shown in Figures 4.1 and 4.2.June and September each experienced several rain events, whereas July and most of August weredominated by a lack of rain and warm dry weather.Figure 4.1: Climate variables measured at Alouette Dam over the study period.414.1. Overview of the study periodFigure 4.2: Open site longwave radiation (top), direct (middle) and diffuse (bottom) shortwaveradiation calculated at the open site climate station on Alouette Dam.Figure 4.3 shows reservoir discharge and streamflow at the WSC gauge for the study period(2013). The daily maximum, minimum, and mean are shown for each day from 1998-2012, whichcorresponds to the period when a minimum release has been provided throughout the summer (BCHydro, 2009). The reservoir discharge was similar to the historical mean during the periods ofspillway and LLO release. Streamflow at the WSC gauge was similar to the mean historical rate;however, streamflow during the dry summer period was roughly 0.3 m3/s lower on average.424.1. Overview of the study periodFigure 4.3: Daily reservoir discharge (top) and streamflow at the WSC gauge (bottom) during thestudy period. Maximum, minimum, and mean flow rates are shown for the period after 1998 whena minimum reservoir discharge provision was initiated.434.2. Time series analysis of stream temperature4.2 Time series analysis of stream temperature4.2.1 Overview of stream temperature time seriesTemperature time series are presented for Tidbits T1, T14, T22, and T26 for the following periods:(1) spillway open and weak seiching signal (Figure 4.4), (2) spillway closed and weak seiching signal(Figure 4.5), (3) spillway closed and strong seiching signal (Figure 4.6), and (4) warm releases andweak seiching signal (Figure 4.7). These periods exhibited distinct temperature patterns, whichappear to be governed by the release temperature (T1). General trends included a strengtheningof the diurnal signal moving downstream, with a more angular diurnal signal at the mouth (T26).Sporadic oscillations with less than 12-hr periods occurred at T1 when the LLO was active, whichdissipated in late August. These signals are assumed to be the result of a metalimnetic withdrawallocation as discussed in more detail in the wavelet analysis section .444.2. Time series analysis of stream temperatureFigure 4.4: Stream temperature time series at T1 (LLO), T14 (6.6 km D/S), T22 (13.9 km D/S),and T26 (Alouette mouth) from June 6 - June 13, 2013 (period 1).Period 1Minimal temperature variation was observed at T1 during period 1 because the LLO temper-ature logger was located away from the spillway and thus not directly influenced by the releases.Both day- and night-time temperatures were higher at T14 than T1, and temperatures at T22 wereslightly higher than at T14. There was warming between T22 and T26 during both day and night.The lack of warming between T14 and T22 is likely to result from shading of solar radiation by454.2. Time series analysis of stream temperaturedenser riparian vegetation and possibly the effect of cool inflows. The increased warming to T26is likely related to less riparian vegetation and greater stream widths, resulting in greater solarradiation exposure.Figure 4.5: Stream temperature time series at T1 (LLO), T14 (6.6 km D/S), T22 (13.9 km D/S),and T26 (Alouette mouth) from June 23 - June 30, 2013 (period 2).Period 2The LLO was opened and spillway closed on June 14, 2013. Summer stratification had not fully464.2. Time series analysis of stream temperatureformed in Alouette Lake by period 2 and there was no evidence of a strong diurnal or semi-diurnalsignal at T1. The signals with less than 12-hr periods have amplitudes up to 4 \u00E2\u0097\u00A6C, and appear tointerfere with the diurnal signal at T14 and T22. A lack of diurnal signal at T14 and T22 couldalso be explained by variable weather and flow rate. The diurnal variation at T26 was minimal andmore box-shaped than sinusoidal, with a minor semi-diurnal signal apparent at the beginning andend of the period.Figure 4.6: Stream temperature time series at T1 (LLO), T14 (6.6 km D/S), T22 (13.9 km D/S),and T26 (Alouette mouth) from July 22 - July 30, 2013 (period 3).474.2. Time series analysis of stream temperaturePeriod 3The semi-diurnal seiching signal at T1 became more apparent during period 3. The peaktemperatures at T1 were attenuated while moving downstream to T22, while low temperaturesincreased and were translated from early in the night to early in the morning. The diurnal signalbecame dominant by T14. However, the seiching signal apparently interacted with solar heating tocause a rapid increase in temperature in the early morning, followed by a decrease in heating ratein the afternoon, then a stronger peak later in the afternoon. The semi-diurnal seiching peaks atT1 appear to reinforce the diurnal daytime peak. The T26 location features a sharp peak in theafternoon with a rapid decline, and a period of relatively constant temperature through the night.484.2. Time series analysis of stream temperatureFigure 4.7: Stream temperature time series at T1 (LLO), T14 (6.6 km D/S), T22 (13.9 km D/S),and T26 (Alouette mouth) from August 21 - August 29, 2013.Period 4During late August, T1 exhibited little diurnal or semi-diurnal variation. The sub-12-hr oscilla-tions ceased around August 25, at which point the temperature began to vary consistently between19.5-21 \u00E2\u0097\u00A6C. This stability was likely caused by either a lowering reservoir level or dropping thermo-cline, resulting in withdrawal of water from the epilimnion above any seiching influence. T14 hada primarily diurnal signal, but with relatively small amounts of night-time cooling. At the mouth,494.2. Time series analysis of stream temperatureT26 exhibited irregular variations that are presumably related to tidal influences.4.2.2 Seiching signalThe seiching period Ti was calculated as 12.4 \u00C2\u00B1 0.5 h depending on the varying thermocline location.The seiche deflection at the outlet, \u00CE\u00B6i, was calculated as 0.39 m based on a U10 of 3.72 m/s, whichwas the highest wind speed averaged over Ti/4 consecutive hours measured at the dam. The southbasin was considered to be the active lake length because the saddle dividing the two basins wasexpected to interfere with seiching due to its shallow depth (approximately 15 m) and eccentriclocation. The thermocline location was based on monthly thermal profiles (Figure 4.8). Theisotherms show two-layer stratification during summer with a relatively linear profile and weakstratification in the spring. The 21.4 \u00E2\u0097\u00A6C measurement at 11 m depth on September 17 could be aresult of destratification instability or more possibly a measurement error.Figure 4.8: Alouette Lake stratification during the 2013 study period. Isotherms are based onmonthly 1 m interval thermal profiles.504.2. Time series analysis of stream temperatureBand-pass filtering suggested that up to 6 \u00E2\u0097\u00A6C oscillations occurred within the 12-hr signal,which is larger than can be accounted for by the recorded temperature profiles and calculatedseiche deflections. The maximum vertical temperature difference over a 1-m interval was 4.5 \u00E2\u0097\u00A6C,between 7 and 8 m on July 19. Assuming the intake for the LLO was located at the thermocline, thehighest \u00CE\u00B6i estimate (0.70 m) would only correspond to a 3.2 \u00E2\u0097\u00A6C temperature oscillation, which is lessthan observed. This inconsistency likely relates to uncertainty in the wind speed used to calculatethe deflection. Wind speed was measured on the dam 3 m above the lake surface as opposed tothe recommended 10 m, and consequently underestimates the actual value of U10. Furthermore,the thermocline depth and vertical gradients are subject to considerable uncertainty due to thelow temporal resolution of thermal profiles: measurements were only taken once per month, whichcould result in the measurement of different phases of seiche deflection each month.4.2.3 Wavelet analysisLow level outletFigure 4.9 presents wavelet spectra for the LLO temperature time series. Significance levelsare not shown to display more clearly the wavelet power spectra for each frequency and time. Thewhite dashed parabolic line in the bottom of the figure represents the cone of influence, outside ofwhich the boundary conditions distort the result (Torrence and Compo, 1998).514.2. Time series analysis of stream temperatureFigure 4.9: Wavelet transform power spectra for temperature time series T1 located at the LLOoutlet (2013).Flow releases were discharged over the spillway during the beginning of the study period untilJune 14, which is evident from the lack of any signal except for diurnal during this period. The T1logger was hydraulically isolated from the main flow in a slack water pool adjacent to the spillway.After flows were directed to the LLO, a variety of transient signals appeared, with the diurnaland semi-diurnal signals constituting the strongest and most consistent. Signals of 1-12 h periodappeared to occur sporadically, which may be a result of outlet hydraulics, where a variable-density-524.2. Time series analysis of stream temperaturedependent withdrawal layer interacts with the seiche (Stevens and Lawrence, 1997), resulting invariable mixing of the metalimnion.The semi-diurnal signal corresponds with the predicted seiching period (12.4-hr). As antici-pated, the period of strong seiching coincides with periods of stronger winds and stratification ascan be seen in Figures 4.1 and 4.8. The seiching signal appeared consistently from mid-July tomid-August except for a few days near August 1, when wind speeds were lower and likely not strongenough to set up a seiche. The oscillation period of the apparent seiching signal was also consistentfrom mid-July to mid-August, which is expected for a seiching signal given a constant thermoclinedepth.The diurnal signal appeared weak during late June before strengthening in July and tapering offfrom the beginning of August to mid-August. The duration of the strong diurnal signal correspondswith the first half of the summer stratified period. A variation in strong 7-24 h signals occurringin mid-July may be a result of strong stratification forming. In June, a weaker linear stratificationwas observed, which may have resulted in down-mixing of diurnally warmed surface water.Wavelet spectra downstream of the dam and on tributariesWavelet transforms of temperature time series with significance tests for a 95% confidence levelwere performed at eight Alouette River mainstem locations, the mouth of North Alouette River, andCreek (Figure 4.10). The time series differ in length, but each spanned mid-May to late-September.534.2. Time series analysis of stream temperature544.2. Time series analysis of stream temperature554.2. Time series analysis of stream temperatureFigure 4.10: Wavelet transform plots for the (a-h) points of interest along Alouette River, (i)North Alouette River, and (j) Millionaire Creek (2013). The colour scheme represents the spectralpower. Regions with spectral power that is significant at the 95% confidence level are outlined inblack.Three groups of oscillation periods are significant for varying durations at each of the locations,including the diurnal signal, 1-12 h signals, and a semi-diurnal signal. The diurnal and semi-diurnalsignals have continuous stretches of significance for over a month, while the 1-12 h signals pop upand fade frequently. There were a few significant signals with oscillations >24-hr that may havebeen associated with weather systems. The scale of oscillation periods does not include the annualoscillation period.The diurnal signal emerged strongly and clearly at T1 and T2 during July and August, thenweakened and increased in variation from T7 to T12, then regained strength downstream to themouth (T26). The downstream increase in strength of the diurnal signal is expected from the effectof solar radiation received by parcels of water moving downstream.The seiching signal evident from mid-July to mid-August decreased in strength from the LLOdownstream, then unexpectedly regained strength between T22 and the mouth. The duration ofthe semi-diurnal signal also exceeded the duration of Alouette Lake\u00E2\u0080\u0099s two-layer stratification andthe seiching signal, indicating a separate process is likely responsible. This second semi-diurnalprocess was inferred to be the principal lunar M2 tidal signal (12.42 h) (McCarthy, 1992), whichcauses an oscillation in river stage up to T23, including the mouth of North Alouette River. Thisrise in stage interferes with the heating caused by energy inputs with magnitudes independent564.2. Time series analysis of stream temperatureof flow volume and velocity, which creates the semi-diurnal oscillation of temperature. Furtherevidence is provided in the relative absence of a 12-hr signal in the Millionaire Creek wavelet powerspectra (Figure 4.10 j), although there are some unexplained, apparently significant 12-hr signalsthroughout the study period.Cross-wavelet analysisEach of the wavelet analysis temperature time series, except for Millionaire Creek, was comparedto T1 using cross-wavelet analysis. Cross-wavelet transform power spectra with 95% significancelevel are shown in Figure 4.11.574.2. Time series analysis of stream temperatureFigure 4.11: Cross wavelet transform plots for T1 crossed with the points of interest along (a-g)Alouette and (h) North Alouette rivers (2013). The black lines outline regions with 95% significancelevels. The colour scheme represents the spectral power.The results of the cross-wavelet analysis were similar to those of the wavelet analysis. Thediurnal signal was consistently significant with a similar power to T1 at all locations for most ofJuly and August, and parts of May, June, and September. Surprisingly, the 1-12 h signals weresignificantly correlated for parts of the study period at each location. The semi-diurnal signalweakened going downstream, then strengthened after T23, which suggests that the seiching signal584.2. Time series analysis of stream temperaturefaded and was then replaced by the semi-diurnal tidal signal between T22 and T23 (14-15.5 km).4.2.4 Cross-correlationAn example filtered time series, highlighting signals with 9-15 h periods, is provided for T1 (Figure4.12). The filtered series agrees well with the semi-diurnal signal observed in the wavelet analysis,where there is minimal signal prior to June 14, when the LLO was closed, moderate oscillationsof approximately 2 \u00E2\u0097\u00A6C during early summer and autumn, and a strong signal with oscillationsexceeding 5 \u00E2\u0097\u00A6C during the mid-July to mid-August period of strong seiching. The stall in seichingassociated with low wind speeds is evident in the first few days of August.Figure 4.12: Temperature time series T1 filtered with a 9-15 h band-pass filter.The result of band-pass filtering the 9-15 h signals is a smooth oscillation of the semi-diurnalsignal consisting of waves of strong negative and positive correlations with maxima spaced 12-hr apart. Strong autocorrelation of the seiching signal results in similar correlations for each ofthe 12-hr lags observed, with most correlations significant at the 95% confidence level. The lagassociated with the MPC between T1 and each mainstem location consistently corresponded atleast approximately to the estimated travel time from the LLO to each downstream location. Thestrength of the correlations decreased from near 1.0 at T2 to 0.5 at T15. The cross-correlationswith T1 then increased at T22 and T23. A MPC of 0.3-0.4 was still present between T1-T26 andT1-North Alouette River, likely owing to the M2 tidal signal previously discussed.594.2. Time series analysis of stream temperature604.2. Time series analysis of stream temperatureFigure 4.13: Cross-correlation plots between the reservoir outlet temperature and locations along(a-g) Alouette and (h) North Alouette rivers between May and September of 2013. Time serieswere filtered with a 9-15 h period band-pass filter.The translation of the seiching signal downstream was further investigated by cross-correlatingeach mainstem temperature time series with T1 temperature time series and recording the timelag associated with MPC. A plot of these lags as a function of distance downstream is provided inFigure 4.14, where the cross-correlation functions were calculated for the whole study period.614.2. Time series analysis of stream temperatureFigure 4.14: Time lags associated with MPC from cross-correlation analysis compared to traveltimes from estimated velocity during 2013 study period.The time lag associated with the MPC between T1 and each downstream Tidbit location shouldrepresent the travel time of water parcels leaving the dam, assuming no interfering semi-diurnalsignal originating downstream of the dam. The estimated arrival time based on the estimatedsteady velocity of 0.38 m/s from hydraulic geometry agrees reasonably well with the values fromthe cross-correlation analysis. The difference between the first MPC- and velocity-based arrivaltimes can possibly be explained by the plunge pool residence time tr. A detailed survey of theplunge pool was not available; however, a rough estimate of plunge pool average depth of 1-3 mwould correspond to residence times of 0.62 - 1.85 h based on the average LLO discharge rate (Q=2.70 m3/s) and assuming tr = V/Q. The plunge pool is partially offset from the direct streamlinebetween the LLO and the channel head, which suggests the lower tr may be more realistic.The MPC-estimated velocities in the forested reach were 2% higher than those estimated fromthe hydraulic geometry measurements. At the drop in gradient near 6.5 km, the MPC velocities624.2. Time series analysis of stream temperaturedecreased before increasing again around 11 km. The increase after 11 km was unexpected fromthe DEM-derived profile, which could be a result of poor DEM resolution, or larger pools in thesection between 6.5 - 11 km. Using only the MPC-based travel time at T23 (15.5 km), the averagevelocity was calculated as 0.32 m/s. The MPC value at the mouth was associated with the nextday\u00E2\u0080\u0099s LLO temperature, indicating that individual seiching peaks could not be accurately trackedto this location using the MPC method.The tidal cross-correlations (Figure 4.15) provide evidence of the upstream extent of tidal in-fluence. The positive and negative correlations between water surface elevation and temperatureat T26 were approximately 0.6 in absolute value, declining progressively to approximately 0.2 atT22. The weak correlation at T22 is likely a result of the semi-diurnal seiching signal, since tidalinfluence was not observed to extend to T22 during any portion of the study period.634.2. Time series analysis of stream temperatureFigure 4.15: Cross-correlation of river stage in the tidally influenced region with e)T22, f)T23,g)T26, h)and North Alouette during 2013 study period. Reference letters were kept the same asprevious cross-correlation figures.644.3. Field measurements for model parameterization4.3 Field measurements for model parameterization4.3.1 Reach-scale hydrologyMainstem flowThe study period was defined by two reservoir discharge conditions: (1) prior to June 14,reservoir discharge was over the spillway, and (2) after June 14, reservoir discharge was through theLLO (Figure 4.16). The dam outflow was higher during the spillway release, averaging 3.71 m3/s,and lower during the LLO release, with an average of 2.70 m3/s and a minimum of 2.53 m3/s. Onelarge rain event at the end of June resulted in flows of 12.0 m3/s at the WSC gauge. During thedry period between July 15 and August 15, flow at the WSC gauge averaged 2.88 m3/s. The LLOdischarge was regulated by reservoir level, resulting in a minor decline through the dry period wheninflows were minimal and reservoir levels declined.Figure 4.16: Hourly flows at upstream (dam) and downstream (WSC gauge) extents of studyreach during 2013. The release was switched from spillway to LLO on June 14, 2013; ; the timingof this switch is indicated by the dashed vertical line.Mainstem flow measurements collected with the velocity-area method are provided in Figure4.17. If the stream was gaining or neutral through the study reach, streamflow measurementswould all be expected to lie between the flow recorded at the dam and WSC gauge. Most of theflows measured during the spillway discharge period were less than the dam discharge, which couldhave been a result of losses throughout the study reach. However, it is suspected that the spillwaydischarge calculation may have been based on a low resolution calibration. During the LLO release654.3. Field measurements for model parameterizationperiod some measurements were still lower than the dam discharge. The velocity-area method isexpected to have an uncertainty of 5%, which may explain the apparent losses. Only three of thegauging sites had uniform hydraulic conditions and it was expected that large boulders at othersites may have caused errors on the higher end of the expected range. Alternatively, the coarsesubstrate size may have resulted in a portion of the streamflow travelling through the hyporheiczone before emerging somewhere upstream of the WSC gauge. Evidence that the stream may havebeen losing is provided by the tributary inflow measurements in the next section.Figure 4.17: Mainstem streamflow measurements taken between the dam and WSC gauge. Mea-surements are shown as filled circles with error bars representing \u00C2\u00B1 5%.Measured streamflow at each of the gauging sites as a function of distance and drainage areaare provided in Figures 4.18 and 4.19. The general trend for both figures is an increase in dischargemoving downstream until around 11 km or 30 km2, where the rate of increase becomes greater to theWSC gauge on some dates. Flow measurements prior to the spillway closing were not included dueto the suspected inaccuracy of the reservoir discharge data. Measurements indicated a downstreamdrop in flow on some days, which may have been a result of losses by infiltration into the bed orbanks or measurement error.664.3. Field measurements for model parameterizationFigure 4.18: Mainstem flow at each of the gauging sites as a function of distance downstream ofAlouette Dam (2013).674.3. Field measurements for model parameterizationFigure 4.19: Mainstem flow at each of the gauging sites as a function of cumulative drainage area(2013).Tributary inflowsFigure 4.20 shows streamflow for nine of the gauged tributaries where additional flows wereestimated from stage measurements. Flows generally dropped off throughout June and reached arelatively steady baseflow in the dry summer period. During September and October, flow in someof the streams began to increase coinciding with more frequent rain events.684.3. Field measurements for model parameterizationFigure 4.20: Streamflow measurements at nine tributaries with additional flows predicted fromstage measurements. Tributaries a to e were located on the left bank, while f to i were located onthe right bank.Two of the measured tributaries in the kame area (streams b and d) originated at groundwaterseeps emerging from cracks in the till or from buried gravel layers. These groundwater-sourcedstreams had a more steady discharge throughout the study period, declining only slightly throughthe dry period. Only two of the tributaries measured dried out completely, which suggests baseflowin most tributaries may have been from a less transient groundwater source.694.3. Field measurements for model parameterizationPiezometer measurementsTable 4.2 presents vertical head gradients (VHG) measured in the stream bed and horizontalhead gradient (HHG) from the channel banks towards the stream. The greatest upward VHGmeasured was 10.0 mm/m at SMS3 on August 19, while the greatest downward VHG measuredwas -16.1 mm/m at SMS2 on July 16. All HHG measurements indicated flow from banks towardsthe stream, with a maximum HHG of 600 mm/m measured on July 16 at the SMS3 left bankpiezometer. Subsurface monitoring site 2 (SMS2) was located in a reach with lower valley wallsthan the other sites, which could explain the apparent downwelling. The higher incidence ofupwelling at sites 3 (SMS3) and 4 (SMS4) was likely due to the steep valley walls and groundwaterseeps observed in the kame region.Table 4.2: Bed piezometer vertical head gradient and bank piezometer horizontal head gradient.Positive gradients represent flow towards the stream. The depth from ground surface to piezometerbottom is provided in the first row.Date Channel bed (mm/m) Channel bank (mm/m)SMS2 SMS3 SMS4 1L 1R 2L 2R 3L 3R 4L 4RDepth (cm) 31 23 42 38 38 40 55 45 43 52 47Jul 15, 2013 - - - 82.8 30 146.7 - - - - -Jul 16, 2013 -16.1 -15 - - - - - 600 - - -Jul 17, 2013 - - - - - - - - 25 - -Jul 25, 2013 -12.9 -2.5 3.6 72 43.3 170 4 55 390 15 10Jul 31, 2013 -11.3 2.5 1.2 - - - - - - - -Aug 9, 2013 - - - 72 - - - - - - -Aug 10, 2013 -6.5 - 1.2 - - - - - - - -Aug 19, 2013 - 10 2.4 - - - - - - - -Aug 26, 2013 - 5 -2.4 - - - - - - - -Sep 3, 2013 -9.7 5 - - - - - - - - -4.3.2 Shade measurementsA summary of shade transect measurements is provided in Table 4.3. The tallest trees were generallyblack cottonwood (Populus trichocarpa) and western hemlock (Tsuga heterophylla). The immediatechannel banks were typically higher in the residential reach resulting in higher maximum tree heightsabove the water surface. Trees were more offset from the channel bank in shade reaches SR4 andSR5 on average due to the de-forested channel banks. Mean overhang lengths ranged from 3.9 mto 6.5 m, with reach SR3 featuring the greatest mean and maximum overhang due to abundantbigleaf maple (Acer macrophyllum).704.3. Field measurements for model parameterizationTable 4.3: Shade transect measurements for each shade reach. All measurements are in metres.Reach Extent Bank + vegetation height Width to tree trunks Overhang length(km) mean min max mean min max mean min maxSR1 0-2.66 24.6 4.9 65.3 2.5 0 26 3.9 0.2 8.3SR2 2.66-5.68 30.5 8 57.1 4 0 17 3.9 0 9.5SR3 5.68-9.51 27.2 11 52.2 2.5 0 35 6.5 0 12.7SR4 9.51-11.17 30.5 9.4 96.7 7.3 0 40 4.6 0.2 11.8SR5 11.17-13.93 23.6 5.2 72.2 10.9 0 100 4 0 10Figures 4.21(a)-(c) present samples of oblique and hemispherical photographs at transects withmainly coniferous bank vegetation (fv = 0.45), deciduous bank vegetation (fv = 0.22), and partiallyde-forested banks (fv = 0.53). Figure 4.21(a) provides an example of the shorter overhang butdenser vegetation provided by coniferous trees, while Figure 4.21(b) shows a transect with wideoverhang but partially transparent deciduous trees.(a) (b) (c)Figure 4.21: Samples of oblique and hemispherical photographs at transects with (a) mainly conif-erous vegetation, (b) deciduous vegetation, and (c) one de-forested bank. Oblique photographswere taken in the downstream direction and hemispherical photographs were taken in the centreof the channel with the top corresponding to south. The line in the centre of the hemisphericalphotograph in Figure 4.21(c) is a power line that was manually removed for radiation calculations.714.3. Field measurements for model parameterization4.3.3 Channel geometryReach-averaged hydraulic geometry parameters based on wetted channel area are provided in Table4.4. Shade reach 5 featured less geomorphic heterogeneity and consisted of long riffle/runs, whichlikely explains the higher velocity and smaller cross sectional area. Two sample cross sections areprovided for each shade reach in Figure 4.22.Table 4.4: Summary of observed hydraulic geometry values during low flow period for each shadereach including: mean wetted width (W\u00C2\u00AF ), maximum wetted width (Wmax), mean depth (d\u00C2\u00AF), maxi-mum depth (dmax), mean velocity (v\u00C2\u00AF), maximum velocity (vmax), mean wetted cross-sectional area(A\u00C2\u00AF), and maximum wetted cross-sectional area (Amax).Shade Reach W\u00C2\u00AF (m) Wmax (m) d\u00C2\u00AF (m) dmax (m) v\u00C2\u00AF (m/s) vmax (m/s) A\u00C2\u00AF (m2) Amax (m2)SR1 18.83 27.05 0.37 0.55 0.44 0.81 6.75 9.9SR2 19.40 26.7 0.42 0.69 0.36 0.56 8.05 12.89SR3 20.31 26.3 0.42 0.65 0.35 0.43 8.20 10.87SR4 22.15 29.5 0.38 0.57 0.37 0.52 8.19 10.83SR5 20.29 28.05 0.29 0.41 0.56 0.96 5.92 8.09724.3. Field measurements for model parameterizationFigure 4.22: Representative cross sections from shade transects in each shade reach. The toppoint represents the water surface elevation at the time of survey between August 7 and September2, 2013, when flows were steady.4.3.4 Bed and bank temperaturesFigures 4.23 and 4.24 show the bed temperatures along with the stream temperature directly aboveeach bed Tidbit at each of the subsurface monitoring sites during the week of maximum weeklymaximum stream temperature (MWMT). The varying amounts of response to the diurnal and734.3. Field measurements for model parameterizationsemi-diurnal variations likely reflect differences in the advective heat transport by groundwaterdischarge and hyporheic exchange, and also differences in thermal properties of the substrate.Figure 4.25 shows the spot stream bed temperature profiles measured at the bed temperaturestakes. The measurements were all taken during the day when the surface flow was typically warmerthan the bed; however, there were two instances where the surface was cooler than the bed. Allof the stakes located in the centre of the channel (SMS2, SMS3a and SMS4) exhibited seasonalwarming of several degrees at 0.5 m depth, suggesting the absence of strong groundwater influence.Site SMS3b, on the other hand, maintained relatively consistent 0.5-m-depth temperatures throughthe summer, consistent with its near-bank location and the likely influence of shallow groundwaterdischarging laterally from the adjacent slope.Figure 4.23: . Stream bed temperatures (red) at subsurface monitoring sites corresponding to T5,T8, and T10, and stream temperature directly above (black). Temperatures are shown during theweek of the mean weekly maximum temperature.744.3. Field measurements for model parameterizationFigure 4.24: Stream bed temperatures (red) at subsurface monitoring sites corresponding to T13and T20, and stream temperature directly above (black). Temperatures are shown during the weekof the MWMT.754.3. Field measurements for model parameterizationFigure 4.25: Bed temperature profiles measured with bed temperature stakes between July 15 andSeptember 3 at three of the subsurface monitoring sites. SMS3b was located 1 m from the channelbank while the other stakes were in the centre of the channel.Two Tidbits were installed in the bank near locations of groundwater upwelling (Figure 4.26),which were near 14.0 \u00E2\u0097\u00A6C during the dry period. The bank temperatures were similar in temperatureto most of the tributaries measured, which suggests that near-surface groundwater would notprovide a strong cooling source.764.3. Field measurements for model parameterizationFigure 4.26: Floodplain temperatures at a depth of 30 cm at locations where moderate groundwaterupwelling occurred continuously.Temperatures measured at the bottom of the bank piezometers ranged between 11.7-16.6 \u00E2\u0097\u00A6Cand averaged 14.0 \u00E2\u0097\u00A6C over the study period (Table 4.5). The absence of temperatures in the 8-9 \u00E2\u0097\u00A6Crange further suggests that groundwater inflows were dominated by near-surface water as opposedto the cooler water seeping from the kame roughly 70 m from the stream.Table 4.5: Bank piezometer temperature measurements (\u00E2\u0097\u00A6C) at depths of 0.34 - 0.6 m.Date 1L 1R 2L 2R 3L 3R 4L 4RJul 15, 2013 12.9 13.2 12.3 12.7 - - - -Jul 16, 2013 - - - 13.1 11.7 - - -Jul 17, 2013 - 12.7 14.3 - - 13.5 - -Jul 25, 2013 12.9 14.3 13.6 14.7 13.9 13.1 13.8 -Jul 31, 2013 12.8 - - 15.1 13.9 13.8 14.1 14.2Aug 9, 2013 13.2 14.7 - - - - - -Aug 10, 2013 - 14.4 15.4 - - 14.7 15.8 -Aug 19, 2013 - - - 14.7 14.5 15 16.6 -Aug 26, 2013 - - - 14.6 14 14.8 16.6 -Sep 3, 2013 13.8 - 15.1 15.1 14.4 14.4 - -774.3. Field measurements for model parameterization4.3.5 Tributary temperaturesFigure 4.27 shows time series of tributary temperatures on the left and right sides of the stream.All of the left bank tributaries were located within the kame region and were found to be 1-2 \u00E2\u0097\u00A6Ccooler than the right bank streams throughout the study period until late September, when theirtemperatures converged with those on the right bank. The coldest tributary on the left bank wasmeasured at the emergence point of a groundwater spring flowing out of a deep crack in the wall ofthe kame. This cold spring had warmed from approximately 8 \u00E2\u0097\u00A6C to 10.4 \u00E2\u0097\u00A6C on two measurementdays in its 70 m length to the confluence with Alouette River. Most of the tributaries in the kamearea were over 300 m long and only two small springs measured had a temperature similar to thecoldest tributary shown. The right bank tributaries all had similar thermal regimes, except for thecoolest tributary shown in black. The coolest right bank tributary had cobble and boulder sizedsubstrate, which resulted in flows retreating into the hyporheic zone during the dry summer periodbefore emerging within 50 m of the confluence with Alouette. This subsurface flow resulted inmuch lower temperatures, which were only observed at one other tributary along the right bank.The tributaries were all relatively small so the tributary temperature was averaged over the entirestudy period to simplify the model.784.3. Field measurements for model parameterizationFigure 4.27: Right bank and left bank tributary temperatures throughout parts of the studyperiod. The colours indicate different tributaries on each side of the channel. The coldest tributarywas measured at the emergence point of a groundwater seep in the kame area.4.3.6 Electrical conductivityElectrical conductivity measurements collected during salt dilution measurements and spot mea-surements were averaged for each Tributary (Figure 4.28). Conductivities in the kame region trib-utaries were on average higher than the right bank and upstream of the kame, providing furtherevidence of a deeper groundwater source.794.3. Field measurements for model parameterizationFigure 4.28: Tributary electrical conductivities averaged at each site over the study period. Eachtributary was categorized into one of the runoff regions with similar runoff characteristics describedin Section 4.4.4.3.7 Above-stream microclimateFigures 4.29 to 4.31 show the difference between above-stream and open site climate variables. Itwas suspected that the above-stream Kestrel loggers were experiencing heating from solar radiation,which compromised the accuracy of air temperature and relative humidity measurements. Sites 5,6, and 8 were located under full canopy cover and were therefore used to fit values for the rest ofthe period. These sites indicated that above-stream microclimate would be slightly cooler than theopen site during the hottest periods.804.3. Field measurements for model parameterizationFigure 4.29: Air temperature at each of the above-stream riparian climate stations plotted againstair temperature measured at the open site.814.3. Field measurements for model parameterizationFigure 4.30: Relative humidity at each of the above-stream riparian climate stations plottedagainst relative humidity measured at the open site.824.4. Data processing for stream temperature modellingFigure 4.31: Wind speed at each of the above-stream climate stations plotted against wind speedmeasured at the open site.4.4 Data processing for stream temperature modelling4.4.1 Flow distributionMainstem flowsFlow at the WSC gauge was smoothed using the loess{stats} function in R (R Core Team,2013). Figure 4.32 shows the pre-processed flow at WSC, where hourly fluctuations of 0.3 m3/s werecommon, especially during August. These fluctuations were assumed to reflect stage measurementerrors at the WSC gauge. Because there was often less than 0.3 m3/s of inflows throughout theentire study reach, it was difficult to quantify inflows and determine whether they were groundwateror surface runoff. The bottom half of Figure 4.32 shows a close-up of early August, illustrating howthe smoother maintains the diurnal evapotranspiration signal while removing hourly variations. Thesmoother also reduced the frequency of periods in which dam outflow was greater than streamflow834.4. Data processing for stream temperature modellingat the WSC gauge.Figure 4.32: Streamflow at the WSC gauge, with raw data and loess smoothed data to removehourly noise.The proportion of total study reach inflows (RQ) as a function of distance downstream ofthe dam are illustrated in Figure 4.33. Values of RQ >1 and <0 were omitted following theassumption that streamflow was increasing in the downstream direction. The accuracy of the flowmeasurements was not sufficient to determine a temporally varying relation between streamflow844.4. Data processing for stream temperature modellingand distance downstream of the dam, so RQ values were averaged over the study period at thethree most frequently repeated gauging sites. A fourth flow measurement site at the 1.64 km mark,which agreed with the trend, is also shown. Discharge was found to increase rapidly in the first2.66 km, followed by a gently increasing plateau to 11.17 km, with a sharp increase to the WSCgauge at 13.93 km.Figure 4.33: RQ values at each flow gauging site on each flow measurement day. The line connectsthe average RQ values for each of the three most frequently repeated gauging sites.Tributary inflowsContinuous records of flow from the non-gauged tributaries were estimated from statisticalrelations between unit tributary discharge (flow divided by drainage area) and Q2 at the WSC gauge(Figure 4.34). The square of discharge was used as a predictor variable to address nonlinearity inthe relation. Separate regressions were fitted for each of three regions with similar relations betweenflow and drainage area. The three regions were the entire right bank (RB), the left bank upstreamof the kame (LBUS), and the left bank in the kame region (LBDS). No measurements were taken854.4. Data processing for stream temperature modellingon the left bank downstream of the kame region, so the average relation from the other regions(combined) was applied. The streamflow for Tributary b plotted well above the fitted relation.This lack of fit was assumed to be a result of inaccurate estimation of the drainage area; therefore,it was removed from the LBDS regression.Figure 4.34: Relations between tributary flow divided by drainage area and the square of dischargeat the WSC gauge. Separate relations are shown for the entire right bank (RB), the left bankupstream of the kame (LBUS), the left bank in the kame region (LBDS), and all data combined.Groundwater inflowsThe groundwater contribution along the reach was assumed to equal the difference between themainstem flow and the tributary inflow at each of the three main streamflow gauging sites (Figure4.35). During the dry period of the summer the estimated surface runoff accounted for more flowthan was apparently entering the stream throughout the study reach. This condition was treatedas a flow loss during which water was assumed to permanently leave the study reach by infiltrationinto the bed and/or banks.864.4. Data processing for stream temperature modellingFigure 4.35: Distribution of cumulative inflows as overland or groundwater at the WSC gaugelocation.Hyporheic exchangeA hyporheic exchange of 2.26\u00C3\u009710\u00E2\u0088\u00925 m2/s was applied in the model, which was based on thehighest average VHG (-11.3 mm/m at SMS2), and a hydraulic conductivity of 10\u00E2\u0088\u00924 m/s estimatedwith the methods of Freeze and Cherry (1979). The average study reach channel width (20 m) wasused to calculate total exchange per longitudinal metre. No evidence of hyporheic outflow into thefloodplain was observed with the bank piezometers, which all showed HHGs towards the channel.The highest VHG was applied to determine if hyporheic exchange would have an impact using thehigher range of estimates. Despite using the upper estimate, hyporheic exchange was confirmed toprovide minimal influence.4.4.2 Stream surface radiation modellingNet radiationModelled net radiation based on the hemispherical photographs agreed well with the measuredvalues (Figure 4.36). The slight temporal misalignment between the modelled and measured valueswas likely a result of an offset between the locations of hemispherical photographs and the netradiometer. Both locations were in areas with dominantly deciduous bank vegetation, where smallgaps in the foliage can be seen as spikes in net radiation throughout the day.874.4. Data processing for stream temperature modellingFigure 4.36: Net radiation measured with the net radiometer and modelled with hemisphericalphotos taken at the same locations as the net radiometer: (a) a location in the upper forested reachon August 9, 2013, and (b) one in the residential reach on September 12, 2013.Shortwave radiationComparisons of shortwave radiation at the open site and modelled at the stream surface areprovided in Figure 4.37. The shortwave radiation modelled with hemispherical photographs wasused to calibrate the geometric shade model. Eight sample transects are provided to illustratethe agreement between models. Transects xs8, xs10, xs23, and xs56 demonstrate how the modelaccurately blocks direct shortwave at the appropriate time regardless of the sun side of the stream.Many of the other eight transects not shown had poor alignment between the hemispherical mod-elled and geometrically modelled values as a result of local variations in tree height or local gapsin the canopy, which were averaged out in the geometric model.884.4. Data processing for stream temperature modellingFigure 4.37: Modelled incident shortwave radiation on a clear sky day (July 27, 2013) at selectshade transects where hemispherical photos were taken. The open grey circles represent the opensite direct shortwave. The black line represents the modelled direct shortwave radiation usinghemispherical photos. The coloured circles represent the geometrically modelled direct shortwavewith colours indicating which side of the stream the sun was over.Examples of incident shortwave radiation hitting the stream for each shade reach on clear skyand partially cloudy days are shown in Figure 4.38. The forested reaches (R1 and R2) experiencedhigher peaks due to the more north-south orientation of the stream. The partially residential894.4. Data processing for stream temperature modellingreaches (R3-R5) experienced more radiation over the course of the clear sky day, likely as a resultof thinner vegetation and lack of high valley walls. On the partially cloudy day, the radiationwas higher in the north-south reaches around midday, and higher in the east-west reaches duringmorning and late afternoon. R3 had lower radiation than the other residential reaches as a resultof longer canopy overhang lengths and leaf density. R5 was exposed to the most daily shortwaveradiation mainly due to thinner bank vegetation and more de-forested banks.Figure 4.38: Reach-averaged direct shortwave radiation reaching the channel for each shade reach(R1-R5) on (a) a clear sky day (July 27, 2013), and (b) a partially cloudy day (June 21, 2013).The open site direct shortwave radiation is shown in grey.Longwave and diffuse shortwaveModelled sky view factors from the geometric view factor model are shown in Figure 4.39.Initial runs resulted in a much lower view factor than modelled from the hemispherical photos.Therefore, multiplication factors between 0.2 and 1 for the overhang length and vegetation heightwere run with 0.05 increments to find the best agreement with hemispherical-photograph-modelled904.4. Data processing for stream temperature modellingview factors. This adjustment seemed reasonable considering the partial transparency of bothoverhanging and bank vegetation, and since the effective height of the bank trees obstructing thesky view was likely lower than the total height.The combination of overhang and tree height reduction factors that resulted in the lowest rootmean squared error (0.098) were applied to each of the shade transects. View factors ranged from0.01 when there was nearly complete canopy closure to 0.65. A mean view factor of 0.29 wascalculated for the full study reach. The transects with de-forested banks generally had higher viewfactors.Figure 4.39: Modelled sky view factors for each shade transect (left) and view factor calibrationfit (right) using hemispherical modelled sky view factors. View factors for transects with partiallyde-forested banks are shown with red circles.4.4.3 Above-stream microclimateKestrel sites 5, 6, and 8 were well-shaded and assumed to provide accurate representations of theabove-stream microclimate; they were therefore used to fit regressions using open site climate vari-ables as predictors. The fitted values for above-stream climate were used in the stream temperaturemodel and assumed uniform over the full study reach. Regression summaries are provided in Ta-ble 4.6. Vapour pressure (ea) was fitted instead of relative humidity since it also accounts for airtemperature. Above-stream wind speed only needed fitting for the period when the Kestrel was atsite 5 due to spiders jamming the anemometer.914.4. Data processing for stream temperature modellingFigure 4.40: Linear regression fits of above-stream climate station variables as a function of opensite climate variables. The red lines represent 1:1 fits.Table 4.6: Summary of regressions relating above-stream to open-site climate variables.Variable Predictor variables (open site) R2 Residual standard error p-valueTa Ta, T 2a 0.90 1.14\u00E2\u0097\u00A6C 2.2E\u00E2\u0088\u009216ea ea 0.64 0.20 kPa 2.2E\u00E2\u0088\u009216WS WS, Ta 0.21 0.34 m/s 2.2E\u00E2\u0088\u0092164.4.4 Bed heat conductionBed temperatures modelled at each 5 cm increment using the finite difference approach are shownfor the week of MWMT in Figures 4.41 and 4.42. The bed heat conduction rates averaged overeach flow reach are provided in Figure 4.43 for the week of MWMT. The direction of heat exchangealternated from downward in the day to upward at night, except in FR1 and partially in FR2 when924.4. Data processing for stream temperature modellingthe seiching signal resulted in a semi-diurnal oscillation. The cooling periods were slightly strongerthan warming in each reach.Figure 4.41: Modelled bed temperatures at Tidbit T20 during week of MWMT using streamtemperature and bed temperature at 0.35 m. This location represents an area of downwelling flow.934.4. Data processing for stream temperature modellingFigure 4.42: Modelled bed temperatures at Tidbit T5 during week of MWMT using streamtemperature and bed temperature at 0.3 m. This location represents an area of upwelling orminimal flow.944.4. Data processing for stream temperature modellingFigure 4.43: Modelled bed heat conduction in each of the four flow reaches during the week ofMWMT.4.4.5 Hydraulic geometry with increased flowFigure 4.44 shows the modelled hydraulic geometry values. The top left plot shows the agreementbetween the target and modelled flows, where a modelled flow within \u00C2\u00B11% of the target flow wasused to calculate hydraulic geometry variables. The modelled values for each individual transectare shown in grey to indicate the variation throughout the study reach and uncertainty associatedwith the assumption of uniformity. Island sections were excluded from the calculations to simplifythe calculations.The variable power equation developed by Ferguson (2007) behaved as expected. Specifically,each of the width, depth, and velocity increased with increasing flow with a decreasing rate. Velocitywas found to increase more rapidly when flows are lower, due to a declining influence of bed frictionand decelerating channel area increase as flows increase.954.5. Stream temperature model evaluationFigure 4.44: Modelled changes to hydraulic geometry variables with increasing flow. Grey linesrepresent each transect interpolated between each streamflow increment, while black lines representthe study reach average interpolated between each streamflow increment (circles).4.5 Stream temperature model evaluationThe stream temperature model described in the analysis section was implemented using each es-timated heat flux and flow time series. The input variables and parameters were adjusted to thehigher, middle, or lower estimate for each of the fluxes to achieve the best agreement between mod-964.5. Stream temperature model evaluationelled and observed temperatures. Particular focus was placed on finding good agreement during thewarm dry period of most concern. The test runs with each of the high, medium, and low estimatesof each variable are shown in Figure 4.45. The errors as a function of distance are reasonablyevenly distributed around the best fit for the parcels arriving at the WSC gauge in the eveningand the morning, which suggests that there could be multiple parameter combinations that wouldprovide good estimates. One of the parameter combinations featuring an equation that predictedhigher incoming longwave radiation resulted in much higher predicted temperatures. The ensembleof model runs shown with the observed T22 temperature at the bottom of Figure 4.45 indicates areasonable fit for each of the combinations, with an average spread of 2.60 \u00E2\u0097\u00A6C between the highestand lowest prediction.974.5. Stream temperature model evaluationFigure 4.45: Model uncertainty analysis during the week of MWMT. The temperature variationof a parcel of water as it flowed downstream is shown for a parcel cooler (top) or warmer (middle)upon arrival at the WSC gauge. Temperature variation at the WSC gauge as function of time isshown on the bottom. The grey lines represent model simulations using high and low estimates ofeach flux, while the black lines represent the best estimate.The parameter combination producing the best fit from the uncertainty analysis was applied tothe rest of the study period for the full study reach (Figure 4.46), the upper half from the dam toT14 (Figure 4.47), and the lower half from T14 to T22 (Figure 4.48). For the full study reach run,984.5. Stream temperature model evaluationthe errors are reasonably well distributed around zero, with some autocorrelation. Over-predictionoften occurred in the day while under-prediction often occurred in the night. Each of the runs alsoover-predicted following the large rain event in late June. The gradual recovery from overestimationin the following few weeks suggests that a higher groundwater discharge could have occurred thattapered off with the dry season. Additionally, higher flows would have been associated with higherdepths and thus a lower temperature response to a given net energy exchange. Unfortunately,measured mainstem and tributary flows were not available in the week following the storm to testthis hypothesis. The under-prediction at the end of the study period appears somewhat correlatedwith flow. However, the effects of tequisous (leaf fall) were not included in the shading model;therefore, the model may have incorporated too much shade resulting in underestimation. Theerror plots of the partial runs indicate that the model slightly over-predicts temperatures at T14,and slightly under-predicts when run from T14 to T22. This pattern of error could be due to animproper characterization of multiple fluxes in isolation or combination. Root mean squared errors(RMSE) for the entire study period for each of the three model lengths are provided in Table 4.7.Figure 4.46: Modelled stream temperature for full study period. Water parcels were released atthe end of the plunge pool and simulated to the WSC gauge.994.5. Stream temperature model evaluationFigure 4.47: Modelled stream temperature from plunge pool to midway at location of T14.Figure 4.48: Modelled stream temperature from midway to end of study reach.1004.6. Modelled heat fluxesTable 4.7: Summary of model error for runs including the full study reach, the upper half, and thelower half.Run length RMSE (\u00E2\u0097\u00A6C) RMSE of daily maxima (\u00E2\u0097\u00A6C)dam to T22 (full study reach) 0.54 0.50dam to T14 0.38 0.60T14 to T22 0.40 0.424.6 Modelled heat fluxesFigure 4.49 shows the total heat exchanged for each heat flux for each parcel over the week ofMWMT. The water temperature at the location of model output (the WSC gauge) is shown forreference. The energy exchanges were summed over the parcel\u00E2\u0080\u0099s travel time, which results inthe maximum contribution of shortwave radiation occurring at the end of the day. The diffuse anddirect components of shortwave radiation provided the only consistently positive heat fluxes. Directshortwave radiation is generally the strongest flux, except on cloudy days when diffuse shortwavecan be more powerful. The dip in direct shortwave can be explained by the sun\u00E2\u0080\u0099s azimuth aligningwith the lower reaches of the stream in the morning, then aligning with the upper reaches duringthe afternoon. Parcels that reached the lower east-west reach when the sun was transitioning toa north-south alignment experienced relatively less direct solar radiation. Net longwave radiationwas consistently negative due to the higher emissivity of water compared to the atmosphere. Theeffect of lower direct shortwave on August 3 can be seen in the lower peak stream temperature.Sensible heat exchange was generally positive during the day and negative at night, while latentheat exchange was consistently negative but greater in absolute magnitude for parcels arriving inthe evening. The effect of bed heat conduction oscillated from negative in the day to positive inthe night. Sensible heat exchange and bed heat conduction were the only two fluxes that oscillatedbetween positive and negative with the diurnal cycle.Tributary inflows provided a cooling source throughout the week with the strongest coolingcorresponding to parcels travelling throughout the day. Groundwater flows were assumed absentduring the period due to evidence of flow losses. Hyporheic exchange provided limited coolingduring the day and limited heating during the night. All of the fluxes except for shortwave radiationare partially dependent on stream temperature; therefore, each flux shows some degree of diurnaloscillation.1014.6. Modelled heat fluxesFigure 4.49: Net heat exchanged by each modelled heat flux as a parcel of water flows throughthe study reach. Stream temperature recorded at the downstream end of the reach is shown in theupper panel, surface fluxes are shown in the middle, and bed heat conduction and the advectivefluxes are shown on the bottom.Figure 4.50 shows the relative percentage of total negative and total positive fluxes associatedwith each term in the energy budget. For parcels arriving during the night, bed heat conductionaccounted for over 80% of the positive terms. The dominant negative term for parcels arriving inthe night was the longwave radiation emitted by the stream. The dominant negative term acting1024.6. Modelled heat fluxesover the course of the day was the latent heat flux. The tributary inflows provided a negative fluxthroughout the week and were most important for the parcels arriving just after midnight. Forparcels arriving during the day, the direct shortwave radiation made up about 2/3 of the dailypositive exchange while diffuse shortwave made up approximately 1/3. Diffuse shortwave radiationaccounted for nearly all of the daily positive flux on cloudy days. Sensible heat became a moreimportant positive term later in the day, often reaching 10%.Figure 4.50: Modelled heat exchange fluxes as a percentage of total positive (middle) and totalnegative (bottom) during the week of MWMT.1034.7. Management scenarios4.7 Management scenariosTable 4.8 provides a summary of temperature changes at the WSC gauge associated with the man-agement scenarios. Figures 4.51 to 4.54 show the modelled daily mean and maximum temperaturesat the WSC gauge based on each of the management scenarios for the full period. The bank re-forestation scenario only resulted in modest reductions of peak temperatures equivalent to 0.18 \u00E2\u0097\u00A6Cover the full period. The daily mean temperature was similar to the existing conditions, with adifference of only -0.08 \u00E2\u0097\u00A6C. The 5-m release depth scenario resulted in decreases in daily mean andmaximum temperatures prior to June 14 due to the surface release over the spillway. Following theswitch to LLO release, The 5-m release depth increased the temperature, resulting in an increase indaily mean and maximum temperature of 0.92 \u00E2\u0097\u00A6C and 0.5 \u00E2\u0097\u00A6C respectively. Temperatures decreaseprogressively with depth of intake for the LLO, with a maximum reduction in daily mean andmaximum of -4.47 \u00E2\u0097\u00A6C and -4.61 \u00E2\u0097\u00A6C with a 20 m release depth. Increased discharge results in atemperature decrease when the release temperature is relatively cool compared to the equilibriumtemperature of the stream, and temperature increases when the release temperature is relativelywarm. This can be seen in Figures 4.52 and 4.54, where a reversal from cooling to warming oc-curs in early August. The effect of progressively increasing flow appears relatively linear; however,there is evidence of a lower bound during spring and early summer, and an upper bound duringAugust and early September. The effect of increased discharge along with changes in release depthconsistently strengthen the effects observed with each scenario on its own.1044.7. Management scenariosFigure 4.51: Daily mean temperatures modelled at the WSC gauge with re-forested banks (top)and changes to release depth (bottom). Release depth, d, is indicated as metres from the watersurface.1054.7. Management scenariosFigure 4.52: Daily mean temperatures modelled at the WSC gauge with changes to reservoirdischarge (top) and combinations of discharge and release depth changes (bottom). Release depth,d, is indicated as metres from the water surface, and discharge, Q, is in m3/s.1064.7. Management scenariosFigure 4.53: Daily maximum temperatures modelled at the WSC gauge with re-forested banks(top) and changes to release depth (bottom). Release depth, d, is indicated as metres from thewater surface.1074.7. Management scenariosFigure 4.54: Daily maximum temperatures modelled at the WSC gauge with changes to reservoirdischarge (top) and combinations of discharge and release depth changes (bottom). Release depth,d, is indicated as metres from the water surface, and discharge, Q, is in m3/s.During the week of MWMT, a reduction in peak temperature achieved by the re-forestationscenario can be seen for a few hours (Figure 4.56). Temperatures are consistently increased by a 5-m release depth, and decreased by 15- and 20-m release depths. The release scenarios were selectedfar enough from the thermocline to avoid the effects of the seiche. The discharge increase scenariosfeature a similar pattern to the full period daily maximum and mean with an additional effect ofadvanced diurnal signal associated with the increased velocity. The maximum temperatures wereexacerbated on August 4 and 10, but dampened on August 5-9. This could be a result of theseiching stage, where the diurnal peak aligns with either the warm or cool phase of the seiche.The effect of 20-m release depth lowers the temperature progressively with higher discharge, andthe effect of 5 m release depth raises the temperature increasingly with higher discharge. Table4.8 includes a summary of the change in mean and maximum temperatures associated with eachscenario during the week of MWMT.1084.7. Management scenariosFigure 4.55: Modelled temperatures at the WSC gauge during the week of MWMT with re-forestedbanks (top) and changes to release depth (bottom). Release depth, d, is indicated as metres fromthe water surface.1094.7. Management scenariosFigure 4.56: Modelled temperatures at the WSC gauge during the week of MWMT with changesto reservoir discharge (top) and combinations of discharge and release depth changes (bottom).Release depth, d, is indicated as metres from the water surface, and discharge, Q, is in m3/s.1104.7. Management scenariosTable 4.8: Model results for management scenarios.Scenario Full period MWMT weekMean \u00E2\u0088\u0086Tw Daily max \u00E2\u0088\u0086Tw Mean \u00E2\u0088\u0086Tw Daily max \u00E2\u0088\u0086Tw(\u00E2\u0097\u00A6C) (\u00E2\u0097\u00A6C) (\u00E2\u0097\u00A6C) (\u00E2\u0097\u00A6C)Re-forested banks -0.08 -0.18 -0.13 -0.45Release depth = 5 m 0.92 0.5 1.65 1.11Release depth = 15 m -3.54 -3.73 -4.31 -4.48Release depth = 20 m -4.47 -4.61 -5.57 -5.66Q + 1 m3/s 0.03 0.08 0.04 -0.06Q + 2 m3/s 0.05 0.08 0.07 -0.09Q + 4 m3/s 0.1 0.07 0.11 -0.05Q + 8 m3/s 0.13 0.02 0.13 -0.2Q + 1 m3/s & depth = 20 m -5.58 -5.63 -6.92 -6.86Q + 1 m3/s & depth = 5 m 1.23 0.83 2.15 1.66Q + 2 m3/s & depth = 20 m -6.19 -6.29 -7.72 -7.73Q + 2 m3/s & depth = 5 m 1.43 0.99 2.46 1.94111Chapter 5DiscussionThis chapter begins with a discussion of general stream temperature patterns observed during thestudy period (Section 5.1). The components of the stream temperature model are then discussedin terms of the challenges of heat flux characterization and suitability of the temporal and spatialresolution (Section 5.2). The model performance and error patterns are then discussed (Section5.3), followed by a discussion of management scenario effectiveness (Section 5.4). The potentialeffects of climate change on thermal regime are then discussed (Section 5.5), followed by a generaldiscussion of the implications of the studies findings on aquatic habitat (Section 5.6).5.1 Stream temperature patternsWith a typical diurnally driven system, a parcel of water released in the morning would be ex-pected to progressively warm as it flows downstream. Progressive warming occurred in June;however, during July and August the opposite often occurred as a result of high epilimnetic releasetemperatures.The wavelet analysis revealed a semi-diurnal signal in each of the stream temperature records.The temporal variation in the strength of this signal was consistent with the presence of factorsthat are conducive to internal seiching (i.e., strong winds and thermal stratification), and theperiod that would be expected for internal seiching in Alouette Lake. Downstream of 15 km,stream temperature appeared to be influenced by semi-diurnal tidal variations in Pitt River waterlevels which obscured the seiching signal. Therefore, the influence of impoundment on AlouetteRiver extends at least 15 km downstream before potential recovery to pre-dam conditions. Otherrecovery lengths cited in the literature have been greater than 40 km (Sinokrot et al., 1995; Oldenand Naiman, 2010); however, those rivers have generally been much larger than Alouette.The thermal regime of reservoir releases was found to be complex as a result of varying releaserate and withdrawal layer, similar to the observations of Webb and Walling (1997). As the reservoirlevel declined through the summer, releases became progressively more epilimnetic, as suggested byNull et al. (2013). The release composition was composed of multiple layers; however, in contrastto Hamblin and McAdam (2003), the variations in composition depended more on internal seichingand declining water level than discharge.The strength of the semi-diurnal signal decreased downstream of the dam, consistent with the1125.1. Stream temperature patternsincreasing influence of diurnal heating/cooling, which became dominant 5.6 km downstream, andnon-diurnal processes such as cloud passage and air mass changes. An increase in diurnal variationoccurred during the period of lower flows; however, this may have been more related to seasonalmeteorological variations rather than to decreases in thermal inertia.The average diurnal variation was strongest approximately 12-hr travel time downstream duringboth release conditions, except at the outlet and downstream end of the plunge pool during thesteady LLO release. This result was similar to the findings of Polehn and Kinsel (1997) andLowney (2000), despite the variations in release temperature. At the location of 12-hr travel timedownstream of the reservoir, average diurnal variation of 2.85 \u00E2\u0097\u00A6C was observed while releases wereover the spillway, and 3.4 \u00E2\u0097\u00A6C while releases were steady and through the LLO. The average diurnalvariation of release temperature during the period of LLO release was 5.6 \u00E2\u0097\u00A6C, which was 3.1 \u00E2\u0097\u00A6Cgreater than during spillway release. This difference suggests that internal seiching can causegreater diurnal scale fluctuations of release temperatures at depth than at the surface. However,seasonal variations must also be taken into consideration.Release temperature and rate were relatively steady for four days at the end of August whenreleases appeared to be strictly epilimnetic. During this period, the pattern described by Polehnand Kinsel (1997) and Lowney (2000) was more apparent, with diurnal oscillation twice as strong12-hr travel time downstream compared to the end of the plunge pool. The stream temperatureat the mouth of Alouette was estimated to be approximately 24-hr downstream. Similar to thefindings of Lowney (2000), there was less diurnal variation at this location then at the 12-hr duringboth release conditions; however, the diurnal variability at this location was suspected to be moredriven by the tidally induced variations in stream depth.The kame region was expected to provide cooling due to observations of cold groundwater seepsand higher tributary discharge. However, an average increase in peak temperatures of 0.2 \u00E2\u0097\u00A6C wasfound between T5 (upstream of the kame area) and T14 (downstream of the kame area) over thefull study period. When T5 temperatures were over 18 \u00E2\u0097\u00A6C, peak temperatures were only decreasedby 0.2 \u00E2\u0097\u00A6C on average through the kame region. Therefore, the effect of kame region groundwaterinflows was likely minor. During the period of mostly epilimnetic release in late August, coolingtypically occurred in the downstream direction; however, on some days there was minimal coolingthrough the kame region despite high release temperatures, which suggests the more variable surfaceexchanges may have been the dominant cooling mechanism.1135.2. Model components5.2 Model components5.2.1 RadiationShortwave radiation was the most important heating term during the day, while longwave radiationwas the most important heat loss term at night. During the calibration of the shortwave andview factor models, it was found that representing the canopy as partially transparent, similar toChen et al. (1998a), Cristea and Burges (2010), and Li et al. (2012), was more appropriate thanassuming an opaque canopy similar to Knudson (2012) and Moore et al. (2014). The net radiationmeasurements also confirmed that measurable direct shortwave radiation was passing through partsof the canopy. Transects where opaque bank vegetation would block all direct shortwave fromhitting the stream were found to experience up to half the total daily open site direct shortwave.Reaches with thin riparian forest zones were also found to provide less shade despite the presenceof more than one row of trees.Preliminary runs of the stream temperature model resulted in over-prediction of daily peaktemperatures. After reviewing the thresholds that were used to determine the gap fractions fromthe hemispherical photographs, it was concluded that thresholds were set too low. Ideally, at leastone day of net radiation would have been recorded each transect where hemispherical photographswere taken. The two locations where net radiation was measured were not adequate to establish thefull range of appropriate thresholds throughout the study reach. Additional confounding variablesincluded the stage of tequisous of bank vegetation, as well as time of day and cloud density whenhemispherical photographs were taken, as noted by Leach and Moore (2010) .Reduction factors applied to the view factor model may have been required due to partialtransparency of the vegetation, tequisous (leaf fall) having begun, or the complexity of tree shapesnot being accurately represented by the geometric model, particularly for deciduous species. Somereaches had more bigleaf maple (Acer macrophyllum), which provided greater overhang and leafarea compared to the cottonwood and alder species. As a result of the complexity of the canopygeometry and its spatial variability, it was difficult to assign lengths and heights for the geometricmodel that properly averaged local conditions. This spatial complexity was compounded by leaffall and the associated temporal variation in radiation transmission through the canopy.Similar to the findings of Li et al. (2012), modelling incident radiation on Alouette River wouldhave been difficult without extensive transect surveys due to spatial variations in canopy andundergrowth overhang, foliage transparency, tree height, and the occurrence of island sections.The agreement between incident shortwave modelled with the hemispherical photographs and thegeometric model (r2 = 0.81) was similar to that achieved by Li et al. (2012) (r2 = 0.90) for theeight transects shown in Figure 4.37. The lesser agreement in this study is suspected to be a resultof using the same canopy transparency factor for each transect, or variations in vegetation heightand width characteristics that were inaccurately averaged at each transect.1145.2. Model components5.2.2 Mainstem flow and inflowsTributary inflows accounted for between 20 to 30% of cooling during the week of MWMT. Duringdry-weather periods, the total increase in streamflow between the dam and the WSC gauge wasabout 0.3 m3/s, only about double the expected error associated with manual flow measurements,which was estimated to be \u00C2\u00B1 0.135 m3/s (Turnipseed and Sauer, 2010). The accuracies of reser-voir discharge and WSC flows were unknown. Many of the manual streamflow measurements inthe mainstem suggested flow losses, which were difficult to confirm given the high magnitude ofmeasurement uncertainty. However, the range of inflow estimates tested in the uncertainty analysisresulted in a change in predicted temperatures of less than 0.3 \u00E2\u0097\u00A6C for the entire week of MWMT.Delineating drainage areas to predict inflows from ungauged tributaries was subject to un-certainty due to the relatively low resolution of the DEM. Field and satellite image observationsof tributary locations often disagreed with those predicted by the DEM-based channel networkmodel. Additionally, complex surficial geology featuring lenses and inter-bedding of glacio-fluvialsandy gravel, glacio-lacustrine stoney-silt, and ice-contact deposits (Geological Survey of Canada,1980) may have resulted in underground flow paths that were not captured by the surface topog-raphy. However, there was little evidence of significant groundwater discharge into the mainstemduring the dry period. This absence may have been a result of the steep valley walls causing ground-water to discharge into tributaries before reaching the mainstem. Groundwater inflows may havebeen higher following the large rain event in late June, which would be consistent with the over-prediction of stream temperatures following the rain events. However, no tributary or mainstemflow measurements were available in the 10 days following the rain event to verify this hypothesis.The sharp increase in mainstem flow in the last 3 km of the study reach was unexpected becausetributary inflow was not found to be higher in this region. Therefore, groundwater discharge waslikely the cause of the flow increase. One reasonable hypothesis is that subsurface flow could becoming from North Alouette River, which approaches Alouette River near the 12 km mark andcomes within 190 m. The coarse substrate and Fraser River alluvium could provide a preferentialflow path between the two streams. A survey of the water surface and bed elevation of NorthAlouette River relative to Alouette River could help assess this hypothesis. The inflows could alsobe due to the drop in mainstem gradient causing upwelling of hyporheic water or groundwaterflowing from the adjacent mountain slopes running alongside the historical Fraser River floodplain.Evidence of hyporheic downwelling was provided by observations of negative in-stream VHG,strong diurnal signals in some bed temperature time series, and apparent losses in flow. Ourability to quantify hyporheic exchange was limited by the small number of piezometer sites, theinstallation of only one bed temperature sensor at each site, and the fact that hydraulic conductivitywas estimated based on visual grain size observation rather than by piezometer tests. Ideally, moresubsurface monitoring sites should have been installed. Installation of multiple bed temperature1155.2. Model componentssensors at each site would have allowed vertical advective exchanges through the stream bed tobe estimated by inverse modelling of bed temperature profiles (Constantz, 2008; Herb and Stefan,2011; Voytek et al., 2014).5.2.3 Turbulent exchangesLatent heat of evaporation accounted for 32% of cooling with a higher percentage in the day reachingas high as 60%. Sensible heat was less important, and never accounted for more than 18% of coolingduring the week of MWMT. The latent heat flux was an important heat sink, similar to the findingsof Benner and Beschta (2000) and Hannah et al. (2008), and appeared to be more important thanin other studies (Story et al., 2003; Moore et al., 2005b; Leach and Moore, 2010). However, thoselatter three studies focused on cooler streams, which would have lower surface vapour pressuresand thus a decreased tendency for evaporation.Accurately characterizing above-stream air temperature and associated vapour pressure provedto be difficult using the Kestrel loggers. Overheating of the Kestrel loggers may have been due toa combination of relatively low wind speed and lack of a radiation shield. The Kestrel anemometerappeared to operate properly through most of the study period. Overall, air temperature was lowerand vapour pressure higher over the stream than measured at the open site. Above-stream windspeed was highly variable, with frequent gusts and periods of no movement, but was generally lowerthan measured at the open site. These differences between above-stream and open-site weatherconditions are consistent with previous studies (Chen et al., 1993, 1995; Spittlehouse et al., 2004;Leach and Moore, 2010; Guenther et al., 2012).This study confirmed the need to characterize climate variables directly above the stream forthe purposes of calculating the sensible and latent heat exchanges, rather than using data from aland-based station. Regression fits to predict these values have been developed for multiple studies(Chen et al., 1993, 1995; Spittlehouse et al., 2004; Leach and Moore, 2010; Guenther et al., 2012).However, these regressions are site-specific, and it would be useful to develop generalized relationsthat explicitly account for differences in stream width, bank height and riparian vegetation amongsites.5.2.4 Bed heat fluxesBed heat conduction was found to moderate stream temperature, accounting for over 90% of night-time warming and up to 33% of daytime cooling. Bed heat conduction calculations relied on singleTidbits buried 30-35 cm in the stream bed at each of the five subsurface monitoring sites. Theabsence of temperature time series at intermediate depths proved problematic, since spot bed tem-perature measurements at 5, 20, and 40 cm often indicated non-linear thermal gradients. Spotmeasurements of vertical profiles were only measured during the day and could not be used to1165.2. Model componentscharacterize diurnal variations. The finite difference approach employed in this study assumes noadvective heat exchange. This assumption may explain the concave temperature gradient mod-elled at T20, where a more linear temperature gradient would be expected considering the closealignment between bed and surface temperatures in Figure 4.24. As mentioned above, installationof multiple bed temperature sensors at each site would allow both advective and conductive heattransfers in the bed to be estimated by inverse modelling.5.2.5 Inflow and groundwater temperatureThe difference between temperatures in the kame region tributaries and the others was only approx-imately 2 \u00E2\u0097\u00A6C, so tributary inflow temperatures were averaged over the entire site. The uncertaintyanalysis confirmed that this assumption was reasonable, since running the model with the lowestand highest tributary temperatures resulted in a change in predicted temperature of less than 0.3\u00E2\u0097\u00A6C for the entire week of MWMT.Electrical conductivity is often used as an indication of inflow source, with deeper groundwa-ter sources often having lower temperatures and higher conductivities (Leach and Moore, 2011).Although some kame area inflow sources had higher conductivity and lower temperatures, they typ-ically flowed overland long enough to warm up before reaching Alouette River. Only one locationwas identified where a cool groundwater seep emerged within 5 m of the stream. The temperaturesof bank seeps were typically similar to the average tributary temperature and likely heated as thewater slowly descended the bank. Due to the low or absent groundwater inflow during the dryperiod, characterizing groundwater temperature with a high degree of accuracy was not consideredimportant. The bank Tidbits also indicated that shallow groundwater was generally as warm asthe tributaries. Spot measurements of apparent groundwater seeps were typically between 13-15\u00E2\u0097\u00A6C.The uncertainty analysis included model runs using the coolest streambed Tidbit data as thegroundwater inflow temperature, and 25% more surface runoff with the lowest tributary tempera-ture. Both of these runs resulted in temperatures within 0.3 \u00E2\u0097\u00A6C of the baseline model during theweek of MWMT. However, it was assumed that tributary subcatchments drained into the mainstemvia open channel flow, which may not have been the case. There were many subcatchments whereoverland flow was not observed and drainage could have been in the form of cooler groundwater.Groundwater inflow temperatures are often estimated by adding 1 to 2 \u00E2\u0097\u00A6C to the mean annualair temperature (Sinokrot et al., 1995). This approximation may be more suitable for lower gradientlandscapes where groundwater retention is longer. The mean annual air temperature at AlouetteDam from 1960 to 2012 was 9.1 \u00E2\u0097\u00A6C, which would result in a groundwater temperature estimate of10.1 to 11.1 \u00E2\u0097\u00A6C. These estimates are approximately 2 - 5 \u00E2\u0097\u00A6C lower than the observed near-surfacegroundwater temperatures in the Alouette catchment.1175.3. Model performance5.2.6 Hydraulic geometryVelocity was set at the mean value based on stream gauging on the mainstem, and was assumedto be steady and uniform. The cross-correlation analysis suggested a decrease in velocity betweenroughly 8 to 12 km. The range of back-calculated Manning\u00E2\u0080\u0099s roughness coefficients (0.03-0.35)indicates non-uniform hydraulics. Observed changes in morphology, including the shift betweenreaches of cascade-riffles and long pool-runs to dominantly riffle-runs, also indicate the likelihoodof a non-uniform velocity. The steady flow assumption was reasonable during extended periods ofdry weather, when streamflow varied slowly through time. Streamflow only varied substantiallyduring and after the late June rain event, when temperatures were not high enough to be of concern.Modelling a 10% lower and higher velocity as part of the uncertainty analysis resulted in upto 1.16 \u00E2\u0097\u00A6C deviations from the base simulation. Therefore, the model is reasonably sensitive tovelocity.5.3 Model performance5.3.1 Model performance compared to similar studiesThe RMSE values of 0.54 \u00E2\u0097\u00A6C for all predictions and 0.50 \u00E2\u0097\u00A6C for daily maxima are similar to thosereported in other applications of process based models, which ranged from 0.03 to 1.12 \u00E2\u0097\u00A6C (Stefanand Sinokrot, 1993; Sinokrot et al., 1995; Foreman et al., 2001; Gooseff et al., 2005; Yearsley,2009; Cristea and Burges, 2010). Most salmonid thermal thresholds are reported to the nearest \u00E2\u0097\u00A6C(Carter and Region, 2005), which suggests that prediction errors of less than 1 \u00E2\u0097\u00A6C are suitable foridentifying risks.The model was solved using a 10-min time step. A time step of 1 min was also tested, with resultsrarely differing by more than 0.15 \u00E2\u0097\u00A6C from the 10-min results. Many predictive stream temperaturemodels have employed an hourly (Gooseff et al., 2005; Yearsley, 2009; Cristea and Burges, 2010)or daily (Stefan and Sinokrot, 1993; Sinokrot et al., 1995) time step. Those resolutions would havebeen too coarse to capture peak temperatures that could exceed acute lethal limits in AlouetteRiver.5.3.2 Error trendsTwo consistent error trends were observed, a synoptic scale pattern and a diurnal pattern, discussedbelow.Synoptic scaleThe synoptic scale error trend featured an average under-prediction during late June, followedby a progressive shift from over-prediction in early July to under-prediction in late September.1185.3. Model performanceThe error trend roughly corresponds to the variations in streamflow at the WSC gauge, whichrose with a rain event in late June and then declined until mid-July. It is suspected that theover-prediction in July was a result of mis-characterization of either latent heat, inflow rate, inflowtemperature, hyporheic exchange, or velocity, and the under-prediction in September was relatedto either radiation or latent heat.The error during higher flows could be due to unsteady velocity as discussed in section 5.2.However, the peak of positive error lags behind the drop in flow, which should be temporallyaligned if the errors were related to hydraulic geometry.A possible explanation of the over-prediction trend during late June is an underestimation ofthe rate of cooler groundwater inflow. Most of the over-prediction occurred in the upper halfof the study reach, where the high relief and kame feature would suggest higher groundwaterinflows. Streamflow measurements did not indicate an increase in streamflow in this region, so itwas difficult to justify the inclusion of higher groundwater inflows. However, it is possible thatthe stream gained flow laterally from tributaries and groundwater at the same time as losing flowvertically through the bed. However, the model run with higher and cooler inflows only resultedin temperature declines of less than 0.3 \u00E2\u0097\u00A6C, and resulted in under-prediction during the week ofMWMT. Therefore, a temporally varying RQ value may have been necessary that accounted for ahigher groundwater component earlier in the study period.It is unlikely that lateral hyporheic exchange was occurring at a high rate, due to high relief,low sinuosity, and positive head gradient towards the stream indicated by bank piezometers. Mis-characterized vertical hyporheic exchange is more probable, given the coarse substrate and relativelack of fines downstream of the dam.The general error trend also roughly followed the seasonal pattern of air temperature, and couldtherefore be related to under-predicted vapour pressure gradients and associated losses throughlatent heat. This could be a result of the three Kestrel sites used to fit air temperature notaccurately representing the mean above-stream air temperature of the study reach.The errors did not follow the rise in streamflow that occurred in September. The under-prediction during this period, beginning in mid-August, is expected to be related to the shademodel not accounting for tequisous, which would result in a greater amount of shortwave radiationduring the day.Diurnal scaleDaytime over-prediction and nighttime under-prediction occurred throughout most of the studyperiod. This pattern of diurnal errors propagated through the entire study period, which wouldsuggest a seasonal scale influence such as mis-specified groundwater temperature or inflow rate.The magnitude of the errors decreased throughout the summer with the drop in mainstem flow,consistent with the hypothesis that the error was related to groundwater influences. The accuracy1195.4. Management scenariosof Alouette Dam outflow data was suspect, and may have been lower than the values reported,which would result in greater inflow rates that could have been in the form of groundwater.Hyporheic exchange moderates diurnal variations in stream temperature (Moore et al., 2005a;Arrigoni et al., 2008) and is, therefore, another possible cause of the diurnal scale errors in thisstudy. Unfortunately, the field measurements were not sufficient to quantify this term with a highlevel of confidence.Improper view factor characterization is another possible cause. It may have led to over-prediction of diffuse shortwave during the day and under-prediction of incident longwave radiationat night.5.4 Management scenarios5.4.1 Bank re-forestationThe reduction in MWMT of 0.45 \u00E2\u0097\u00A6C achieved by re-forestation is within the uncertainty limits ofthe model. However, the consistent reduction on clear days and negligible reduction on the cloudydays suggest that the relative decrease is realistic. The decrease in peak temperatures associatedwith 4% re-vegetation in the study reach suggests that greater temperature reductions could beachieved in the agricultural reach, where de-forestation is more severe. The east-west orientationin the agricultural reach also means that re-foresting only the south bank would be necessary tomitigate peak shortwave radiation. Diminishing shade from riparian vegetation typically occurswith downstream increases in channel width (Vannote et al., 1980; Cristea and Burges, 2010); how-ever, the channel width remained relatively constant through the study reach. In the agriculturalreach, an approximate doubling of channel width occurs, which could impact the success of bankre-forestation. Establishment of fully grown trees along the agricultural reach banks could pro-vide shading of approximately 25% of the channel at solar noon for red alder (Alnus rubra) andapproximately 50% for black cottonwood (Populus trichocarpa).Planting trees in the residential reach at the same density as in the forested reach could alsoprovide further improvements, since some of the residential reach was lined with only one row oftrees, which allowed transmission of shortwave radiation. Some of the de-forested banks featuredoverhanging herbaceous vegetation that provided shading. An additional management option couldbe to restore the shorter shrub and herbaceous bank vegetation on otherwise un-vegetated banks.The effects of re-vegetation on above-stream microclimate and the associated fluxes were notconsidered. The effect could either be a reduction in stream temperatures from lowering air tem-perature, or increased stream temperatures from decreasing wind and evaporation rates. There-forestation option could be difficult to implement, as landowners may not be receptive to re-foresting their river frontage.1205.4. Management scenarios5.4.2 Reservoir release regimeThe management scenarios including hypolimnetic release of water from the reservoir resulted inthe greatest temperature reductions, with MWMT reductions at the WSC gauge consistently 4 \u00E2\u0097\u00A6Cgreater than any of the other scenarios. The period when the lake was stratified (mid-July to mid-September) corresponded with the greatest release temperatures, which reinforces the importanceof withdrawal depth. Stratification regime and stream temperatures both depend on seasonalvariations of air temperature, streamflow, and solar radiation, and therefore the same periods maybe the hottest in similar but unregulated systems.Increased release rates resulted in a moderate reduction of MWMT, likely a result of increasedthermal inertia. Increased hourly variation was also predicted during peak hours of each day, likelyassociated with the increased strength of the seiching signal, which amplifies or dampens the peaktemperatures on different days.Stream temperature has been found to be inversely correlated with flow rate in unregulatedrivers (Hockey et al., 1982; Sinokrot and Gulliver, 2000). This observation was not found to holdfor the regulated Alouette system, where a coupling of higher flows and high release temperatureswas found to cause greater downstream warming. This result is of high importance to dam operatorssince conventional understanding of an inverse relation between flow and temperature could lead tomitigation efforts inadvertently making conditions worse. An awareness of the release temperatureand reservoir stratification regime are therefore essential in guiding any mitigation actions.A spillway discharge scenario was not modelled since high-resolution reservoir surface temper-ature data were not available due to logger theft. Gooseff et al. (2005) found that surface releasesfrom a relatively shallow reservoir warmed river temperature by approximately 3 \u00E2\u0097\u00A6C. Risley et al.(2010) similarly predicted an increase of 2-3 \u00E2\u0097\u00A6C following dam removal and loss of hypolimneticrelease. These studies highlight the cooling effects of hypolimnetic release on downstream temper-ature.Webb and Walling (1997) observed highly variable stratification and release regimes that affecteddownstream temperature and would add uncertainty when predicting acute thermal threats tosalmonids. High annual variability in thermal stratification has been observed in Alouette Lake,which further reinforces the need to base release decisions on a knowledge of current stratificationconditions. Alouette Lake currently features a relatively simplified release regime, since the releasewater is not subject to the typical fluctuations associated with releases used for power generation.Determining release rate is also a socioeconomic issue that depends on recreational use, powerdemand for the surrounding area, and power pricing where it can be more lucrative for hydro-powerproducers to store or release water at different times depending on the market. Changes to reservoirrelease regimes should also consider the potential impacts on other water quality parameters, suchas dissolved oxygen, nutrients, pollutants, and sediment loading that can vary with reservoir depth1215.5. Potential effects of climate change(Caissie, 2006; Olden and Naiman, 2010).5.4.3 Other management optionsAdditional warming mitigation possibilities exist in the residential and agricultural reaches thatwere not modelled. The application of best management practices (BMPs) and low impact de-velopment (LID) are often implemented to achieve water quality targets including thermal regime(Walker, 1995; Krause et al., 2004). Risley et al. (2010) found that groundwater withdrawals in aregulated river could cause warming of less than 0.5 \u00E2\u0097\u00A6C during the summer, which is in the orderof the modelled reductions from reforesting Alouette River\u00E2\u0080\u0099s residential reach. Examples of BMPsthat may successfully reduce Alouette River stream temperature would be (1) to vegetate roadsideditches in the residential area to provide shading and associated cooling of inflows, (2) to limit theamount of impermeable pavement to allow groundwater recharge, and (3) to install bottom-releasestormwater management ponds deep enough to stratify. The scarcity of summer rainfall could limitthe success of these BMPs; however, BMPs (2) and (3) would help cool inflows and spread out theirduration, thereby providing local thermal refugia for resident and migrating fish as described byMcCullough (1999).The reservoir system studied by Webb and Walling (1997) featured a springflow component ofcompensation flows. A similar component could be installed at Alouette Dam to allow a portionof the releases to flow through a cooling medium before reaching Alouette River.5.5 Potential effects of climate changeThe impact of climate change on Alouette River is expected to be mostly related to increasing airtemperature because Alouette River has a regulated streamflow regime that could be maintainedat a constant level. However, summers in south coastal British Columbia are projected to becomewarmer and drier, which would result in reduced lateral inflows. Tributary inflow constitutedapproximately 20% of the cooling terms during the week of MWMT and groundwater was assumedabsent. Since the tributaries were likely groundwater fed, a warming of groundwater as predictedby Sinokrot et al. (1995) could affect their cooling influence.5.6 Implications for aquatic habitatSince 2005, a surface release over Alouette Dam has allowed the out-migration of juvenile kokaneethat have been returning during July and August in numbers between 5 and 103 per year (Borick-Cunningham, 2013). One of the management concerns identified has been whether Alouette River\u00E2\u0080\u0099sthermal regime would be suitable to support a re-established sockeye run (FWCP, 2011). Restoring1225.6. Implications for aquatic habitatfish passage past Alouette Dam is also being considered, which depends in part on the ongoingmonitoring studies being facilitated by BC Hydro (FWCP, 2011).The wider and slower agricultural reach with minimal riparian vegetation resulted in muchhigher temperatures at the mouth. During July and August, when sockeye typically begin entry,outlet temperatures often exceeded 25 \u00E2\u0097\u00A6C, the lethal limit for most salmonid species (Carter andRegion, 2005), and reached as high as 26.4 \u00E2\u0097\u00A6C. Temperatures as low as 19 \u00E2\u0097\u00A6C have been foundto prevent river entry of migrating salmonids (Carter and Region, 2005). Entry avoidance couldexplain the relatively poor 2013 sockeye return (Borick-Cunningham, 2013). Thermal thresholdsfor the historical Alouette sockeye stock are unknown; however, upper lethal temperatures havebeen found to begin as low as 20.8 \u00E2\u0097\u00A6C for other Fraser River sockeye stocks (Eliason et al., 2011).A general guideline proposed by EPA. (2003) suggests that the MWMT should not exceed 16\u00E2\u0097\u00A6C in streams with core juvenile salmonid rearing habitat. This condition was only achieved at theWSC gauge by the scenarios including a hypolimnetic release. To determine if any of the mitigationscenarios could improve thermal conditions between the WSC gauge and the river mouth, it wouldbe necessary to accurately model the tidal influence on stream depth in addition to the processesmodelled upstream.It should also be noted that aquatic ecosystems have complex thermal requirements that dependon stream temperature duration, frequency, magnitude, timing, and rate of change (Olden andNaiman, 2010). Therefore, management decisions should also consider the requirements of thelocal biota rather than absolute temperature decreases as assessed in this study.123Chapter 6ConclusionsThis section summarizes the key findings of the study and provides recommendations for futureresearch on Alouette River and general stream temperature studies.6.1 Key findingsStream temperature patternsThe approximately 12-hr internal seiching signal identified in the reservoir caused release tem-perature oscillations of up to 5 \u00E2\u0097\u00A6C during summer stratified periods. Wavelet and cross-correlationanalysis indicated that the seiching signal propagated at least 15 km before it was obscured byinterference from the M2 tidal signal. Epilimnetic releases during the summer resulted in peaktemperatures often exceeding 20 \u00E2\u0097\u00A6C at the WSC gauge 14 km downstream of the dam. A trendof downstream increase in daytime peak temperatures was found during early June. During theperiod that the reservoir had two-layer stratification, a trend of downstream decrease in peak tem-peratures was observed to the end of the residential reach, followed by an increase through theagricultural reach.Stream temperature modelThe stream temperature model was developed using field measurements of hydrological, meteo-rological, and geomorphic variables that govern the stream heat fluxes. A Lagrangian model similarto that of Leach and Moore (2011) was employed to model temperatures between Alouette Damand the Alouette River WSC gauge 14 km downstream. Using a 10-min time step, an accuracy of0.54 \u00E2\u0097\u00A6C RMSE was achieved for all predictions, and 0.50 \u00E2\u0097\u00A6C RMSE for daily peaks.The model often over-predicted daytime peaks and under-predicted nighttime lows. This waslikely due to mis-characterized mediating fluxes that could have included bed heat conduction, hy-porheic exchange, groundwater inflows, tributary inflows, and sensible and latent heat. The meandaily error had a trend of over-prediction following a large rain event in late June, followed bya progressive decline towards under-prediction by the end of September. It is believed that theover-prediction was related to an under-estimation of cooling inflows when catchment antecedentmoisture was high, and under-prediction when tequisous resulted in greater transmission of short-wave radiation through the canopy. This highlights the importance of characterizing seasonal and1246.1. Key findingsweather system scale patterns in each of the heat fluxes.The relative contribution from each heat flux was summed for each parcel of water from the timeit left the plunge pool to its arrival at the WSC gauge during the week of MWMT. Parcels arrivingin the evening and early night gained most heat through direct and diffuse shortwave radiation,and lost the most heat through latent heat, tributary inflows and bed heat conduction. Parcelsarriving in the early morning gained most heat through bed heat conduction, and lost most heatthrough longwave radiation and tributary inflows.Management scenariosThe effect of re-establishing a natural level of bank vegetation along the de-forested banks wasfound to be minimal, with average peak temperature reductions of 0.18 \u00E2\u0097\u00A6C for the full period, and0.45 \u00E2\u0097\u00A6C during the week of MWMT. These modest reductions result from the fact that only 4%of the bank length along the study reach was de-forested. Potential temperature reductions wouldlikely be higher in the agricultural reach downstream of the study reach, where banks have beenheavily de-forested, similar to the system studied by Cristea and Burges (2010).Modification of the reservoir release depth resulted in the greatest effects on downstream tem-perature. Similar to Gooseff et al. (2005), the result of an epilimnetic release from 5 m depthresulted in a mean downstream increase of 0.92 \u00E2\u0097\u00A6C over the full study period and 1.65 \u00E2\u0097\u00A6C duringthe week of MWMT. Similar to Sinokrot et al. (1995), the hypolimnetic release scenario from 20 mdepth resulted in a mean decrease of 4.47 \u00E2\u0097\u00A6C over the full study period and 5.57 during the weekof MWMT.Increased reservoir discharge resulted in higher thermal inertia with the effect on downstreamtemperatures being highly dependent on the release temperature. Increased velocity resulted inearlier peak temperatures 14 km downstream. Daily mean and peak temperatures increased onaverage during the full period by maxima of 0.13 and 0.08 \u00E2\u0097\u00A6C respectively. Mean temperatureincreased during the week of MWMT by a maximum of 0.13 \u00E2\u0097\u00A6C, while maximum temperature de-creased by up to 0.2 \u00E2\u0097\u00A6C, corresponding to the +8 m3/s scenario. Generally the effects were amplifiedby progressively increasing flows with evidence of upper and lower bounds during the warmest andcoolest periods. The effect of increasing flows along with modified release level amplified the effectof modified release level, with the greatest MWMT reduction (-7.73 \u00E2\u0097\u00A6C) corresponding to +2 m3/sreleased from 20 m depth, and the greatest MWMT increase (2.46 \u00E2\u0097\u00A6C) corresponding to +2 m3/sreleased from 5 m depth. Therefore, this study reinforces the findings of Sinokrot and Gulliver(2000), Webb and Walling (1997), and Risley et al. (2010) that reservoir stratification regime mustbe understood before manipulating release regime.1256.2. Directions for future research6.2 Directions for future researchImprovements to stream temperature modelsPredicting the warming associated with climate change and other anthropogenic influences isimportant to assist in conserving aquatic ecosystems (Nelson and Palmer, 2007). Process-basedstream temperature modelling has proven to be a valuable tool in predicting the effects of watershedmanagement activities on stream thermal regime (Sinokrot et al., 1995; Gooseff et al., 2005; Cristeaand Burges, 2010).Although the model developed in this study generally performed well, a number of terms werenot well constrained. Radiative exchanges were the dominant drivers of warming and cooling, andwere strongly controlled by riparian vegetation. The effects of riparian vegetation are typicallysimulated using geometric models (Moore et al., 2014), which often treat the canopy as opaque.The riparian vegetation along Alouette River had a complex structure that was difficult to representusing simplified geometry, was partially transparent to radiative transfer, and varied through timedue to the significant deciduous component. Further research is required to develop more advancedshade models for these complex riparian canopies.The latent heat flux was an important cooling term at Alouette River. However, there isconsiderable uncertainty in its predicted magnitude due to (a) uncertainty about the applicabilityof the wind function and (b) uncertainty in the above-stream wind speed, air temperature andvapour pressure. Further research is required to develop a parameterization for the sensible andlatent heat fluxes that can be generally applied, taking into account scale effects. In addition, thereis a need for approaches to estimating above-stream microclimate from open-site measurements,taking into account the sheltering and shading effects of the stream banks and riparian vegetation.Management of Alouette River\u00E2\u0080\u0099s thermal regimeEfforts to enhance and restore salmonid populations would be aided by a better understandingof the pre-dam thermal and hydrologic regimes in relation to the thermal requirements of AlouetteRiver\u00E2\u0080\u0099s full species assemblage (Olden and Naiman, 2010). Whereas pre-dam streamflow data exist,there appear to be no pre-dam water temperature data. It would be useful to conduct process-based modelling studies to establish the pre-dam thermal regime. In addition, a regional empiricalmodel, such as that developed by Moore et al. (2013) could be useful.Management scenario modelling could also be conducted with specific thermal targets and anoptimization iteration. For example, release depths at every 1 m increment could be modelled todetermine the ideal release regime. This could guide the development of a variable release depthregime similar to Webb and Walling (1997). Reservoir stratification dynamics were found to greatlyinfluence downstream temperature. Modelling the timing of stratification, thermocline depth, andinternal seiching regime similar to Stevens and Lawrence (1997) and Imberger and Patterson (1981)1266.2. Directions for future researchwould be valuable to more accurately forecast the effects of modified release regime.A closer look at Alouette River\u00E2\u0080\u0099s thermal spatial heterogeneity could help identify thermalrefugia for migrating and resident salmonids (Olden and Naiman, 2010). Lateral, vertical, andlongitudinal heterogeneity investigation could help identify cooler pools, cutoff channels, and localgroundwater and tributary inflows. Modelling stream temperature to the mouth of Alouette Riverat the confluence with Pitt River would be helpful to predict future entry temperatures salmonidsmay face. This would involve modelling the tidal influence as well as the relatively large NorthAlouette River, which joins 3 km upstream of the mouth.Given the need for a better understanding of Alouette River\u00E2\u0080\u0099s thermal regime, further process-based modelling should be conducted. As mentioned above, there is a general need for moreresearch to develop approaches for characterizing stream surface energy exchanges. In addition,in the specific case of Alouette River, there was considerable uncertainty in specifying the reach-scale hydrology, particularly lateral inputs and possible flow losses, and bed heat fluxes. Furtherresearch should put more effort into characterizing these boundary conditions to reduce modellinguncertainties. In particular, the use of piezometers and bed temperature loggers at multiple depthswould allow vertical heat and water fluxes to be estimated by inverse modelling.127ReferencesArrigoni, A. S., Poole, G. C., Mertes, L. A., O\u00E2\u0080\u0099Daniel, S. J., Woessner, W. W., and Thomas, S. A.2008. Buffered, lagged, or cooled? Disentangling hyporheic influences on temperature cycles instream channels. Water Resources Research, 44(9).Balcilar, M. 2007. mFilter: Miscellaneous time series filters. R package version 0.1-3.Baxter, C., Hauer, F. R., and Woessner, W. W. 2003. Measuring groundwater\u00E2\u0080\u0093stream waterexchange: new techniques for installing minipiezometers and estimating hydraulic conductivity.Transactions of the American Fisheries Society, 132(3):493\u00E2\u0080\u0093502.BC Hydro 2009. Alouette Project Water Use Plan. Technical report, BC Hydro.BCRP 2000. Strategic Plan, Volume 2, Watershed Plans, Chapter 6: Alouette River. Technicalreport, Bridge-Coastal Restoration Program.Benner, D. and Beschta, R. 2000. Effects of channel morphology on evaporative heat loss fromarid-land streams. In Proceedings of the AWRA International Conference on Riparian Ecologyand Management in Multi-Land Use Watersheds. Bethesda, MD: American Water ResourcesAssociation.Borick-Cunningham, G. 2013. Alouette water temperature, Program No. ALUMON 3. Technicalreport, BC Hydro.Caissie, D. 2006. The thermal regime of rivers: a review. Freshwater Biology, 51(8):1389\u00E2\u0080\u00931406.Carrivick, J. L., Brown, L. E., Hannah, D. M., and Turner, A. G. 2012. Numerical modelling ofspatio-temporal thermal heterogeneity in a complex river system. Journal of Hydrology,414:491\u00E2\u0080\u0093502.Carter, K. and Region, N. C. 2005. The effects of temperature on steelhead trout, coho salmon,and Chinook salmon biology and function by life stage. Implications for the Klamath Rivertotal maximum daily loads. California Regional Water Quality Control Board. North CoastRegion, Santa Rosa, California.Caryula, C. and Groves, K. 2009. Alouette project water use plan. Technical report, BC Hydro.128ReferencesChapra, S. C. 1997. Surface water-quality modeling. Waveland Press.Chen, J., Franklin, J. F., and Spies, T. A. 1993. Contrasting microclimates among clearcut, edge,and interior of old-growth Douglas-fir forest. Agricultural and Forest Meteorology,63(3):219\u00E2\u0080\u0093237.Chen, J., Franklin, J. F., and Spies, T. A. 1995. Growing-season microclimatic gradients fromclearcut edges into old-growth Douglas-fir forests. Ecological Applications, 5(1):74\u00E2\u0080\u009386.Chen, Y. D., Carsel, R. F., McCutcheon, S. C., and Nutter, W. L. 1998a. Stream temperaturesimulation of forested riparian areas: I. Watershed-scale model development. Journal ofEnvironmental Engineering, 124(4):304\u00E2\u0080\u0093315.Chen, Y. D., McCutcheon, S. C., Norton, D. J., and Nutter, W. L. 1998b. Stream temperaturesimulation of forested riparian areas: II. Model application. Journal of EnvironmentalEngineering, 124(4):316\u00E2\u0080\u0093328.Christiano, L. J. and Fitzgerald, T. J. 2003. The band pass filter. International EconomicReview, 44(2):435\u00E2\u0080\u0093465.Cimmery, V. 2007-2010. SAGA User Guide, updated for SAGA version 2.0.5.Constantz, J. 2008. Heat as a tracer to determine streambed water exchanges. Water ResourcesResearch, 44:W00D10. doi: 10.1029/2008WR006996.Cooke, S. J., Hinch, S. G., Farrell, A. P., Lapointe, M. F., Jones, S. R., Macdonald, J. S.,Patterson, D. A., Healey, M. C., and Van Der Kraak, G. 2004. Abnormal migration timing andhigh en route mortality of sockeye salmon in the Fraser River, British Columbia. Fisheries,29(2):22\u00E2\u0080\u009333.Coutant, C. C. 1999. Perspectives on temperature in the Pacific Northwest\u00E2\u0080\u0099s fresh waters.Technical report, Oak Ridge National Lab., TN (US).Cristea, N. C. and Burges, S. J. 2010. An assessment of the current and future thermal regimes ofthree streams located in the Wenatchee River basin, Washington State: some implications forregional river basin systems. Climatic Change, 102(3-4):493\u00E2\u0080\u0093520.Crossin, G. T., Hinch, S., Cooke, S., Welch, D., Patterson, D., Jones, S., Lotto, A., Leggatt, R.,Mathes, M., Shrimpton, J., Van Der Kraak, G., and Farrell, A. 2008. Exposure to hightemperature influences the behaviour, physiology, and survival of sockeye salmon duringspawning migration. Canadian Journal of Zoology, 86(2):127\u00E2\u0080\u0093140.129ReferencesEaton, B. C. 2006. Bank stability analysis for regime models of vegetated gravel bed rivers. EarthSurface Processes and Landforms, 31(11):1438\u00E2\u0080\u00931444.Edinger, J. E., Duttweiler, D. W., and Geyer, J. C. 1968. The response of water temperatures tometeorological conditions. Water Resources Research, 4(5):1137\u00E2\u0080\u00931143.Eliason, E. J., Clark, T. D., Hague, M. J., Hanson, L. M., Gallagher, Z. S., Jeffries, K. M., Gale,M. K., Patterson, D. A., Hinch, S. G., and Farrell, A. P. 2011. Differences in thermal toleranceamong sockeye salmon populations. Science, 332(6025):109\u00E2\u0080\u0093112.Elo, A.-R., Huttula, T., Peltonen, A., and Virta, J. 1998. The effects of climate change on thetemperature conditions of lakes. Boreal Environment Research, 3:137\u00E2\u0080\u0093150.EPA. 2003. EPA Region 10 Guidance for Pacific Northwest State and Tribal Temperature WaterQuality Standards. EPA 910-B-03-002.Erbs, D., Klein, S., and Duffie, J. 1982. Estimation of the diffuse radiation fraction for hourly,daily and monthly-average global radiation. Solar Energy, 28(4):293\u00E2\u0080\u0093302.Ferguson, R. 2007. Flow resistance equations for gravel-and boulder-bed streams. WaterResources Research, 43:W05427. doi: 10.1029/2006WR005422.Ficke, A. D., Myrick, C. A., and Hansen, L. J. 2007. Potential impacts of global climate changeon freshwater fisheries. Reviews in Fish Biology and Fisheries, 17(4):581\u00E2\u0080\u0093613.Fischer, H. B. 1979. Mixing in inland and coastal waters. Academic Press, San Diego, CA.Foreman, M., Lee, D., Morrison, J., Macdonald, S., Barnes, D., and Williams, I. 2001.Simulations and retrospective analyses of Fraser watershed flows and temperatures.Atmosphere-Ocean, 39(2):89\u00E2\u0080\u0093105.Frazer, G. W., Canham, C., and Lertzman, K. 1999. Gap Light Analyzer (GLA), Version 2.0:Imaging software to extract canopy structure and gap light transmission indices fromtrue-colour fisheye photographs, users manual and program documentation. Simon FraserUniversity, Burnaby, B.C., and the Institute of Ecosytem Studies, Millbrook, N.Y.Freeze, R. A. and Cherry, J. A. 1979. Groundwater. Prentice-Hall, Upper Saddle River, N.J.FWCP 2011. Alouette Watershed Salmonid Action Plan. Technical report, British Columbia Fishand Wildlife Compensation Program.Geological Survey of Canada 1980. Surficial Geology of New Westminister, west of sixthmeridian, British Columbia. Technical report, Geological Survey of Canada.130ReferencesGomi, T., Moore, R. D., and Dhakal, A. S. 2006. Headwater stream temperature response toclear-cut harvesting with different riparian treatments, coastal British Columbia, Canada.Water Resources Research, 42(8):W08437. doi:10.1029/2005WR004162.Gooseff, M. N., Strzepek, K., and Chapra, S. C. 2005. Modeling the potential effects of climatechange on water temperature downstream of a shallow reservoir, Lower Madison River, MT.Climatic Change, 68(3):331\u00E2\u0080\u0093353.Gouhier, T. 2014. biwavelet: Conduct univariate and bivariate wavelet analyses. (Version 0.17.2).Greenbank, J. 2009. Alouette water temperature, Program No. ALUMON 5. Technical report,BC Hydro.Gu, R., Montgomery, S., and Austin, T. A. 1998. Quantifying the effects of stream discharge onsummer river temperature. Hydrological Sciences Journal, 43(6):885\u00E2\u0080\u0093904.Guenther, S., Moore, R., and Gomi, T. 2012. Riparian microclimate and evaporation from acoastal headwater stream, and their response to partial-retention forest harvesting. Agriculturaland Forest Meteorology, 164:1\u00E2\u0080\u00939.GVRD 2004. Kanaka Creek Regional Park Management Plan. Technical report, GreaterVancouver Regional District.Hamblin, P. and McAdam, S. 2003. Impoundment effects on the thermal regimes of KootenayLake, the Arrow Lakes Reservoir and Upper Columbia River. Hydrobiologia, 504(1-3):3\u00E2\u0080\u009319.Hannah, D. M., Malcolm, I. A., Soulsby, C., and Youngson, A. F. 2008. A comparison of forestand moorland stream microclimate, heat exchanges and thermal dynamics. HydrologicalProcesses, 22(7):919\u00E2\u0080\u0093940.Herb, W. R. and Stefan, H. G. 2011. Modified equilibrium temperature models for cold-waterstreams. Water Resources Research, 47(6):W06519. doi: 10.1029/2010WR009586.Hockey, J., Owens, I., and Tapper, N. 1982. Empirical and theoretical models to isolate the effectof discharge on summer water temperatures in the Hurunui River. Journal of Hydrology (NewZealand), 21(1):1\u00E2\u0080\u009312.Hrachowitz, M., Soulsby, C., Imholt, C., Malcolm, I., and Tetzlaff, D. 2010. Thermal regimes in alarge upland salmon river: a simple model to identify the influence of landscape controls andclimate change on maximum temperatures. Hydrological Processes, 24(23):3374\u00E2\u0080\u00933391.Imberger, J. and Patterson, J. 1981. Dynamic Reservoir Simulation Model - DYRESM: 5. In H.Fisher, Transport Models for Inland and Coastal Waters, pages 310\u00E2\u0080\u0093361. Academic Press, NewYork.131ReferencesIqbal, M. 1983. An introduction to solar radiation. Academic Press, Toronto, ON.Isaak, D., Wollrab, S., Horan, D., and Chandler, G. 2012. Climate change effects on stream andriver temperatures across the northwest US from 1980\u00E2\u0080\u00932009 and implications for salmonidfishes. Climatic Change, 113(2):499\u00E2\u0080\u0093524.Isaak, D. J. 2011. Stream temperature monitoring and modeling: Recent advances and new toolsfor managers. Stream Systems Technology Centre: Stream Notes.Jackson, J. 1833. On the Seiches of Lakes. Journal of the Royal Geographical Society of London,3:271\u00E2\u0080\u0093275.Johnson, B. M., Saito, L., Anderson, M. A., Weiss, P., Andre, M., and Fontane, D. G. 2004.Effects of climate and dam operations on reservoir thermal structure. Journal of WaterResources Planning and Management, 130(2):112\u00E2\u0080\u0093122.Knudson, J. M. 2012. The effect of discharge variability on the heat budget and tributary mixingdynamics of a proglacial river. Master\u00E2\u0080\u0099s thesis, University of British Columbia.Krause, C. W., Lockard, B., Newcomb, T. J., Kibler, D., Lohani, V., and Orth, D. J. 2004.Predicting influences of urban development on thermal habitat in a warm water stream.JAWRA Journal of the American Water Resources Association, 40(6):1645\u00E2\u0080\u00931658.Lafreniere, M. and Sharp, M. 2003. Wavelet analysis of inter-annual variability in the runoffregimes of glacial and nival stream catchments, Bow Lake, Alberta. Hydrological Processes,17(6):1093\u00E2\u0080\u00931118.Lane, R. J. 1999. Hydrometric technician career development program, Lesson Package No. 10.1Principles of Discharge Measurement. Technical report, The Water Survey of Canada.Lapham, W. W. 1989. Use of temperature profiles beneath streams to determine rates of verticalground-water flow and vertical hydraulic conductivity. USGS Water-Supply Paper 2337, 35p.Lawrence, D. J., Stewart-Koster, B., Olden, J. D., Ruesch, A. S., Torgersen, C. E., Lawler, J. J.,Butcher, D. P., and Crown, J. K. 2014. The interactive effects of climate change, riparianmanagement, and a nonnative predator on stream-rearing salmon. Ecological Applications,24(4):895\u00E2\u0080\u0093912.Leach, J. A. and Moore, R. D. 2010. Above-stream microclimate and stream surface energyexchanges in a wildfire-disturbed riparian zone. Hydrological Processes, 24(17):2369\u00E2\u0080\u00932381.Leach, J. A. and Moore, R. D. 2011. Stream temperature dynamics in two hydrogeomorphicallydistinct reaches. Hydrological Processes, 25(5):679\u00E2\u0080\u0093690.132ReferencesLessard, J. L. and Hayes, D. B. 2003. Effects of elevated water temperature on fish andmacroinvertebrate communities below small dams. River Research and Applications,19(7):721\u00E2\u0080\u0093732.Li, G., Jackson, C. R., and Kraseski, K. A. 2012. Modeled riparian stream shading: Agreementwith field measurements and sensitivity to riparian conditions. Journal of Hydrology,428:142\u00E2\u0080\u0093151.Liu, Y., Liang, X. S., and Weisberg, R. H. 2007. Rectification of the bias in the wavelet powerspectrum. Journal of Atmospheric and Oceanic Technology, 24(12):2093\u00E2\u0080\u00932102.Lowney, C. L. 2000. Stream temperature variation in regulated rivers: Evidence for a spatialpattern in daily minimum and maximum magnitudes. Water Resources Research,36(10):2947\u00E2\u0080\u00932955.Maheu, A., Caissie, D., St-Hilaire, A., and El-Jabi, N. 2013. River evaporation and correspondingheat fluxes in forested catchments. Hydrological Processes, page doi: 10.1002/hyp.10071.McCarthy, D. D. 1992. IERS Standards for 1992. Technical report, DTIC Document.McCullough, D. A. 1999. A review and synthesis of effects of alterations to the water temperatureregime on freshwater life stages of salmonids, with special reference to Chinook salmon.Technical Report EPA 910-R-99-010. 291 pp., Water Resource Assessment, Columbia RiverInter-Tribal Fish Commission, Portland, OR.Mohseni, O., Erickson, T. R., and Stefan, H. G. 1999. Sensitivity of stream temperatures in theUnited States to air temperatures projected under a global warming scenario. Water ResourcesResearch, 35(12):3723\u00E2\u0080\u00933733.Moore, R. D. 2004. Introduction to salt dilution gauging for streamflow measurement: Part 1.Streamline Watershed Management Bulletin, 7(4):20\u00E2\u0080\u009323.Moore, R. D., Leach, J. A., and Knudson, J. M. 2014. Geometric calculation of view factors forstream surface radiation modelling in the presence of riparian forest. Hydrological Processes,page doi: 10.1002/hyp.9848.Moore, R. D., Nelitz, M., and Parkinson, E. 2013. Empirical modelling of maximum weeklyaverage stream temperature in British Columbia, Canada, to support assessment of fish habitatsuitability. Canadian Water Resources Journal, 38(2):135\u00E2\u0080\u0093147.Moore, R. D., Spittlehouse, D. L., and Story, A. 2005a. Riparian microclimate and streamtemperature response to forest harvesting: a review. JAWRA Journal of the American WaterResources Association, 41(4):813\u00E2\u0080\u0093834.133ReferencesMoore, R. D., Sutherland, P., Gomi, T., and Dhakal, A. 2005b. Thermal regime of a headwaterstream within a clear-cut, coastal British Columbia, Canada. Hydrological Processes,19(13):2591\u00E2\u0080\u00932608.Morrison, J., Quick, M. C., and Foreman, M. G. 2002. Climate change in the Fraser Riverwatershed: flow and temperature projections. Journal of Hydrology, 263(1):230\u00E2\u0080\u0093244.Murdock, T. and Spittlehouse, D. 2011. Selecting and using climate change scenarios for BritishColumbia. Pacific Climate Impacts Consortium, University of Victoria, Victoria, B.C., 39 pp.Nelson, K. C. and Palmer, M. A. 2007. Stream temperature surges under urbanization andclimate change: data, models, and responses. JAWRA Journal of the American WaterResources Association, 43(2):440\u00E2\u0080\u0093452.Null, S. E., Ligare, S. T., and Viers, J. H. 2013. A method to consider whether dams mitigateclimate change effects on stream temperatures. JAWRA Journal of the American WaterResources Association, 49(6):1456\u00E2\u0080\u00931472.Olden, J. D. and Naiman, R. J. 2010. Incorporating thermal regimes into environmental flowsassessments: modifying dam operations to restore freshwater ecosystem integrity. FreshwaterBiology, 55(1):86\u00E2\u0080\u0093107.Pike, R. G., Redding, T. E., Wilford, D. J., Moore, R. D., Ice, G., Reiter, M. L., and Toews, D.A. A. 2010. Climate change effects on watershed processes in British Columbia, Chapter 19.Compendium of forest hydrology and geomorphology in British Columbia. RG Pike, TERedding, RD Moore, RD Winkler, and KD Bladon (editors). BC Ministry of Forests andRange, Forest Science Program, Victoria, BC, and FORREX Forum for Research andExtension in Natural Resources, Kamloops, BC Land Management Handbook 66, pp 699-747.Pilgrim, J. M., Fang, X., and Stefan, H. G. 1998. Stream temperature correlations with airtemperatures in Minnesota: Implications for climate warming. JAWRA Journal of theAmerican Water Resources Association, 34(5):1109\u00E2\u0080\u00931121.Polehn, R. and Kinsel, W. 1997. Transient temperature solution for stream flow from a controlledtemperature source. Water Resources Research, 33(1):261\u00E2\u0080\u0093265.Polehn, R. and Kinsel, W. 2000. Transient temperature solution for a river with distributedinflows. Water Resources Research, 36(3):787\u00E2\u0080\u0093791.Poole, G. C. and Berman, C. H. 2001. An ecological perspective on in-stream temperature:Natural heat dynamics and mechanisms of human-causedthermal degradation. EnvironmentalManagement, 27(6):787\u00E2\u0080\u0093802.134ReferencesPunzet, M., Vo, F., Vo, A., Kynast, E., and Ba\u00C2\u00A8rlund, I. 2012. A Global Approach to Assess thePotential Impact of Climate Change on Stream Water Temperatures and Related In-StreamFirst-Order Decay Rates. Journal of Hydrometeorology, 13(3).Quilty, E. and Moore, R. 2007. Measuring stream temperature. Streamline WatershedManagement Bulletin, 10(2):25\u00E2\u0080\u009330.R Core Team 2013. R: A Language and Environment for Statistical Computing. R Foundation forStatistical Computing, Vienna, Austria.Richards, J., Moore, R. D., and Forrest, A. L. 2012. Late-summer thermal regime of a smallproglacial lake. Hydrological Processes, 26(18):2687\u00E2\u0080\u00932695.Risley, J. C., Constantz, J., Essaid, H., and Rounds, S. 2010. Effects of upstream dams versusgroundwater pumping on stream temperature under varying climate conditions. WaterResources Research, 46:W06517. doi: 10.1029/2009WR008587.Rodenhuis, D., Bennett, K., Werner, A., Murdock, T., and Bronaugh, D. 2007.Hydro-climatology and future climate impacts in British Columbia. Pacific Climate ImpactsConsortium, University of Victoria, Victoria BC, 132 pp.Salmond, J. A. 2005. Wavelet analysis of intermittent turbulence in a very stable nocturnalboundary layer: implications for the vertical mixing of ozone. Boundary-Layer Meteorology,114(3):463\u00E2\u0080\u0093488.Schindler, D. W., Bayley, S. E., Parker, B. R., Beaty, K. G., Cruikshank, D. R., Fee, E. J.,Schindler, E. U., and Stainton, M. P. 1996. The effects of climatic warming on the properties ofboreal lakes and streams at the Experimental Lakes Area, northwestern Ontario. Limnologyand Oceanography, 41:1004\u00E2\u0080\u00931017.Shrestha, R. R., Schnorbus, M. A., Werner, A. T., and Berland, A. J. 2012. Modelling spatial andtemporal variability of hydrologic impacts of climate change in the Fraser River basin, BritishColumbia, Canada. Hydrological Processes, 26(12):1840\u00E2\u0080\u00931860.Silliman, S. E. and Booth, D. F. 1993. Analysis of time-series measurements of sedimenttemperature for identification of gaining vs. losing portions of Juday Creek, Indiana. Journal ofHydrology, 146:131\u00E2\u0080\u0093148.Sinokrot, B., Stefan, H., McCormick, J., and Eaton, J. 1995. Modeling of climate change effectson stream temperatures and fish habitats below dams and near groundwater inputs. ClimaticChange, 30(2):181\u00E2\u0080\u0093200.135ReferencesSinokrot, B. A. and Gulliver, J. S. 2000. In-stream flow impact on river water temperatures.Journal of Hydraulic Research, 38(5):339\u00E2\u0080\u0093349.Sinokrot, B. A. and Stefan, H. G. 1993. Stream temperature dynamics: measurements andmodeling. Water Resources Research, 29(7):2299\u00E2\u0080\u00932312.Spigel, R. H. and Farrant, B. 1984. Selective withdrawal through a point sink and pycnoclineformation in a linearly stratified flow. Journal of Hydraulic Research, 22(1):35\u00E2\u0080\u009351.Spigel, R. H. and Imberger, J. 1980. The classification of mixed-layer dynamics of lakes of smallto medium size. Journal of Physical Oceanography, 10(7):1104\u00E2\u0080\u00931121.Spittlehouse, D. L., Adams, R. S., and Winkler, R. D. 2004. Forest, edge, and openingmicroclimate at Sicamous Creek. British Columbia, Ministry of Forests, Forest Science Program.Stefan, H. and Sinokrot, B. 1993. Projected global climate change impact on water temperaturesin five north central US streams. Climatic Change, 24(4):353\u00E2\u0080\u0093381.Stevens, C. L. and Lawrence, G. A. 1997. Estimation of wind-forced internal seiche amplitudes inlakes and reservoirs, with data from British Columbia, Canada. Aquatic Sciences,59(2):115\u00E2\u0080\u0093134.Story, A. C., Moore, R. D., and Macdonald, J. S. 2003. Stream temperatures in two shadedreaches below cutblocks and logging roads: downstream cooling linked to subsurface hydrology.Canadian Journal of Forest Research, 33(8):1383\u00E2\u0080\u00931396.Stull, R. B. 2000. Meteorology for scientists and engineers: a technical companion book withAhrens\u00E2\u0080\u0099 Meteorology Today. Brooks/Cole.Torrence, C. and Compo, G. P. 1998. A practical guide to wavelet analysis. Bulletin of theAmerican Meteorological Society, 79(1):61\u00E2\u0080\u009378.Turnipseed, D. P. and Sauer, V. B. 2010. Discharge measurements at gaging stations: U.S.Geological Survey Techniques and Methods book 3 chap. A8, 87 p. Technical report, UnitedStates Geological Survey.Vallentyne, J. R. 1957. Principles of modern limnology. American Scientist, 45:218\u00E2\u0080\u0093244.Vannote, R. L., Minshall, G. W., Cummins, K. W., Sedell, J. R., and Cushing, C. E. 1980. Theriver continuum concept. Canadian Journal of Fisheries and Aquatic Sciences, 37(1):130\u00E2\u0080\u0093137.Veleda, D., Montagne, R., and Araujo, M. 2012. Cross-wavelet bias corrected by normalizingscales. Journal of Atmospheric & Oceanic Technology, 29(9):doi:10.1175/JTECH\u00E2\u0080\u0093D\u00E2\u0080\u009311\u00E2\u0080\u009300140.1.136ReferencesVoytek, E. B., Drenkelfuss, A., Day-Lewis, F. D., Healy, R., Lane, J. W., and Werkema, D. 2014.1DTempPro: analyzing temperature profiles for groundwater/surface-water exchange.Groundwater, 52(2):298\u00E2\u0080\u0093302.Walker, W. W. 1995. Design basis for everglades stormwater treatment areas. Water ResourcesBulletin, 31(4):671\u00E2\u0080\u0093685.Webb, B. and Walling, D. 1997. Complex summer water temperature behaviour below a UKregulating reservoir. Regulated Rivers: Research & Management, 13(5):463\u00E2\u0080\u0093477.Webb, B. W., Hannah, D. M., Moore, R. D., Brown, L. E., and Nobilis, F. 2008. Recent advancesin stream and river temperature research. Hydrological Processes, 22(7):902\u00E2\u0080\u0093918.Willson, M. F. and Halupka, K. C. 1995. Anadromous fish as keystone species in vertebratecommunities. Conservation Biology, 9(3):489\u00E2\u0080\u0093497.Yearsley, J. R. 2009. A semi-Lagrangian water temperature model for advection-dominated riversystems. Water Resources Research, 45:W12405. doi: 10.1029/2008WR007629.137"@en . "Thesis/Dissertation"@en . "2014-11"@en . "10.14288/1.0166950"@en . "eng"@en . "Geography"@en . "Vancouver : University of British Columbia Library"@en . "University of British Columbia"@en . "Attribution-NonCommercial-NoDerivs 2.5 Canada"@en . "http://creativecommons.org/licenses/by-nc-nd/2.5/ca/"@en . "Graduate"@en . "Modelling the thermal regime of a regulated coastal British Columbia river and assessing the potential of warming mitigation strategies"@en . "Text"@en . "http://hdl.handle.net/2429/50199"@en .