IDENTIFICATION AND CHARACTERIZATION OF STREAM THERMAL REFUGES AND ASSOCIATED HABITAT USE BY JUVENILE PACIFIC SALMON (ONCORHYNCHUS SPP.) AND STEELHEAD (O. MYKISS) by Thomas Alexander Willms B.N.R.Sc., Thompson Rivers University, 2006 DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY IN NATURAL RESOURCES AND ENVIRONMENTAL STUDIES UNIVERSITY OF NORTHERN BRITISH COLUMBIA October 2024 © Thomas Alexander Willms, 2024 Abstract Cold-water patches can function as thermal refuges for juvenile Pacific salmon (Oncorhynchus spp.) and steelhead (O.mykiss) during periods of high summer stream temperatures. My study uses a three-phased approach to identify and classify thermal refuges in streams, characterize their environmental conditions during critical rearing periods, and monitor diel migration patterns of juvenile salmonids in response to environmental covariates. To identify thermal refuges, I used Remotely Piloted Aircraft System-based thermal infrared imaging. Repeat imaging of my study reach following a major channel-forming flood event in my second field season revealed that the composition of thermal refuges changed significantly, and that abundance of thermal refuges also increased. Specifically, the abundance of cold alcoves, a type of thermal refuge, increased eightfold following this flood. Cold alcoves are created by lateral shifts in stream channel alignments as part of the cut, fill, and avulsion processes associated with unconfined alluvial rivers. My results bring new evidence that flood processes of intact and unconfined floodplains provide an important role in maintenance and recruitment of thermal refuges in streams. In characterizing in situ temperature and groundwater conditions at thermal refuges, I installed temperature loggers and shallow streambed piezometers to collect time-series of temperature data and vertical hydraulic gradient – a measure of groundwater upwelling/downwelling conditions. Changes to these conditions were modelled in response to environmental covariates: mainstream stream temperature, air temperature, and stream discharge. Mainstem stream temperatures and mainstem thermal refuge temperatures were highly sensitive to changes in air temperature, and off-channel thermal refuge temperatures were least sensitive to changes in air temperature. The effects of stream discharge on thermal refuge temperature and ii vertical hydraulic gradient varied by site. Conspicuously, I observed that at some thermal refuges, short-term pulses of stream discharge from dam releases and atmospheric events drove inverse changes in groundwater upwelling. This evidence provides new considerations for regulated rivers that are counter to current understanding and common assumptions regarding hyporheic exchange and stream discharge. To characterize the diel horizontal migration patterns of juvenile salmonids at thermal refuges, I used Passive Integrated Transponder tags and arrays of antennae. Fish were captured, tagged, and released at identified thermal refuges over two summers. I logged and compiled fishspecific detections and determined migration type (i.e., immigration or emigration) based on the sequence of detections across multiple antennae. I fitted models for migration and occupancy at thermal refuges and found that, although fish differentially exploited cold-water patches during periods of high mainstem stream temperatures, stream and thermal refuge temperature were poor predictors of thermal refuge habitat use. Instead, diel horizontal migrations were highly correlated with photoperiod. Fish were found to enter mainstem habitats from thermal refuges at dusk and would make return migrations at dawn. Fish also exhibited thermal refuge site fidelity, and diel horizontal migrations continued into the fall after mainstem stream temperatures had cooled. My work provides new evidence regarding what triggers diel horizontal migrations in stream-dwelling juvenile salmonids, their occupancy at thermal refuge sites, and highlights the complexity of factors in their movement ecology. iii Table of Contents Abstract .............................................................................................................................. ii List of Tables ................................................................................................................ viii List of Figures ................................................................................................................ ix Acknowledgements ....................................................................................................... xii Funding ........................................................................................................................ xiii Prologue .......................................................................................................................... 1 CHAPTER 1: Identification of thermal refuges in streams using remotely piloted aircraft system-based thermal infrared imaging .......................................................................................... 6 Abstract ........................................................................................................................... 7 Introduction ..................................................................................................................... 8 Methods..........................................................................................................................11 Study Location ...........................................................................................................11 Equipment ................................................................................................................. 14 Field Methods ........................................................................................................... 15 Data Processing ......................................................................................................... 17 Results ........................................................................................................................... 19 Density and composition of thermal refuge habitats ................................................ 19 Discussion ..................................................................................................................... 22 iv Effectiveness of methods in identifying thermal refuge habitats .............................. 22 Suggested Best Practices for collection of TIR images and identification of thermal refuge habitats ........................................................................................................... 25 Spatio-temporal dynamics of thermal refuge habitats .............................................. 27 Role of RPAS-based TIR in conservation of thermal refuge habitats in streams ..... 28 CHAPTER 2: Vertical hydraulic gradient and temperature dynamics at stream thermal refuges under changing flow and atmospheric conditions ............................................................ 30 Abstract ......................................................................................................................... 31 Introduction ................................................................................................................... 32 Methods......................................................................................................................... 35 Study Location .......................................................................................................... 35 Site Selection ............................................................................................................ 37 Equipment ................................................................................................................. 38 Data Analysis ............................................................................................................ 41 Results ........................................................................................................................... 42 Site-level summary of temperature and VHG .......................................................... 42 Mainstem temperature response to environmental covariates .................................. 48 Site-level thermal refuge and groundwater temperature response to environmental covariates .................................................................................................................. 49 Stream discharge and VHG dynamics ...................................................................... 53 v Discussion ..................................................................................................................... 56 CHAPTER 3: Diel horizontal migration of stream-dwelling juvenile Pacific salmonids (Oncorhynchus spp.) between mainstem and thermal refuge habitats.......................................... 63 Abstract ......................................................................................................................... 64 Introduction ................................................................................................................... 65 Methods......................................................................................................................... 67 Study Location .......................................................................................................... 67 Site Selection ............................................................................................................ 69 Fish Capture .............................................................................................................. 70 Enumeration, Tagging and Tracking ......................................................................... 70 Temperature Monitoring ........................................................................................... 72 Data Analysis ............................................................................................................ 72 Results ........................................................................................................................... 75 Thermal Refuge Selection and Temperature Summary ............................................ 75 Fish Collection, Tagging, Detection, and Migration Summary ................................ 79 DHM Modelling........................................................................................................ 85 Discussion ..................................................................................................................... 89 Epilogue ........................................................................................................................ 94 References ....................................................................................................................... 100 Appendix A. ................................................................................................................ 120 vi Appendix B. ................................................................................................................ 122 Appendix C. ................................................................................................................ 124 Appendix D. ................................................................................................................ 126 Appendix E. ................................................................................................................ 127 vii List of Tables Table 1.1. Descriptions of thermal refuge habitat classes used in data analysis (Modified from Dugdale et al. 2015). ..................................................................................................................... 19 Table 2.1. Mainstem stream temperature model summary. "ΔT.Air + ΔT.Air(t-1) + ΔQ" was the most parsimonious model, following removal of VHG as an explanatory variable due to correlation issues between Q and VHG at some sites.. ................................................................. 48 Table 2.2. Summary of multi-variate site-level thermal refuge and groundwater temperature flux models by year. β1 represents the air temperature coefficient, β2 represents the lagged (t-1) air temperature coefficient, and β3 represents the mean daily discharge coefficient. ΔAICc represents model improvement from site-level univariate model of air temperature only.. .......................... 52 Table 3.1. Peak (modes) and least (anti-mode) detection times (24 hr.) by site. .......................... 80 viii List of Figures Figure 1.1. 22.4 km study reach (denoted by thickened grey line) located on the Nicola River between its confluences with Coldwater River (upstream) and Spius Creek (downstream), near the city of Merritt, British Columbia, Canada. ............................................................................. 14 Figure 1.2. Example of thermal infrared image showing differences in relative temperature between river mainstem and identified thermal refuge habitat. Sp1 to Sp4 indicate spot measurements of relative temperature and corresponding locations. ........................................... 18 Figure 1.3. Counts of identified thermal refuge habitats by class between 2016 and 2017 TIR surveys. ......................................................................................................................................... 20 Figure 1.4. Image of identified thermal refuge habitat with temperature difference value (-5.4˚C) based on in situ measurement of mainstem habitat temperature at point "Sp2" and thermal refuge habitat temperature at point “Sp1”. Absolute temperatures at Nicola River mainstem and thermal refuge habitat and would be estimated at 19.2˚C and 14.1˚C, respectively. ................................. 23 Figure 1.5. Example of shaded, north-facing stream bank (2017) has the potential to be misidentified as thermal refuge – specifically, lateral seepage. .................................................... 24 Figure 1.6. Evidence of downstream thermal plume can help differentiate potential thermal refuge habitats from thermal artifacts in TIR images. Image shows mixing of cool water from a springbrook thermal refuge. .......................................................................................................... 25 Figure 2.1. Location of study area and individual study sites. Inset map shows the location within the Pacific Northwest region of the United States and Canada. ................................................... 37 Figure 2.2. Streambed groundwater monitoring well diagram (under upwelling conditions), modified from Baxter er al. (2003). .............................................................................................. 39 Figure 2.3. (a) Time-series mean daily stream discharge (Nicola River at Spences Bridge) in relation to (b) mean daily VHG by site in 2017. VHG was highly variable in 2017 but exhibited some general patterns in relation to stream discharge by site. (c) VHG plotted in response to stream discharge by site in 2017. .................................................................................................. 46 Figure 2.4. (a) Time-series mean daily stream discharge (Nicola River at Spences Bridge) in relation to (b) mean daily VHG by site in 2018.. Note the inverse pattern of VHG in response to rapid changes in stream discharge, particularly at Site 2. (c) VHG plotted in response to stream discharge by site in 2018. .............................................................................................................. 47 Figure 2.5. Linear regressions with 95% confidence interval of site-level differences in thermal sensitivity of thermal refuge stilling well temperature (SWTemp) in relation to air temperature (Air Temp) at thermal refuge sites. 2017 data are indicated by solid circles and regression lines, and 2018 data are indicated by open circles and dashed regression lines. All models were highly significant (P = <0.001), except for that of Site 2 in 2018 (P = 0.153) which exhibited no significant relationship between air temperature and thermal refuge temperatures. .................... 50 ix Figure 2.6. Linear regressions with 95% confidence interval of site-level differences in thermal sensitivity of groundwater / piezometer temperature (PzTemp) in relation to air temperature (Air Temp) at thermal refuge sites in 2017. 2017 data are indicated by solid circles and regression lines, and 2018 data are indicated by open circles and dashed regression lines. All models were highly significant (P = <0.001), except for that of Site 2 in 2017 (P = 0.081) which exhibited no significant relationship between air temperature and thermal refuge groundwater temperatures. 51 Figure 2.7. Generalized Additive Model (GAM) of the effects of change in daily stream discharge on change in VHG in 2018, including all sites. Plot includes 95% confident intervals, including standard error of model terms and intercept. Note that although GAMs can be used for modelling non-linear relationships between model variables, this model has an effective degrees of freedom value of 1.003, indicating a linear relationship between ΔQ and ΔVHG. .................. 54 Figure 2.8. Generalized Additive Model of the effects of change in daily stream discharge on change in VHG in 2018 at Site 2 only. Plot includes 95% confident intervals, including standard error of model terms and intercept. Note that this model, fitted to data for Site 2 only, retains the linear relationship between approximately ±1 m3/s ΔQ, but decreases in confidence towards more extreme changes in daily discharge. Changes in discharge at this site were most strongly correlated with inverse changes in groundwater upwelling, according to its VHG. .................... 55 Figure 2.9. Generalized Additive Model of the effects of change in daily stream discharge on change in VHG at Site 3 for years 2017 (a) and 2018 (b). Plot includes 95% confident intervals, including standard error of model terms and intercept. Note that, unlike the Site 2 model included in Figure 2.8, increases in stream discharge are generally associated with increases in VHG. ............................................................................................................................................. 56 Figure 3.1. Location of study area and individual study sites in the Interior of British Columbia, Canada. Inset map shows the location within the Pacific Northwest region of Canada and the United States. ................................................................................................................................ 69 Figure 3.2. Thermal infrared images showing heterogeneity of surface water temperatures between thermal refuge sites and the mainstem Nicola River: (A) Site 1 - at the confluence of Clapperton Creek and the Nicola River; (B) Site 2 – a cool-water outfall from a beaver pond adjacent to the Nicola River; (C) Site 3 – lateral seep of cool ground water and associate offchannel habitat to the Nicola River; and, (D) Site 4 – groundwater-fed alcove created by overbank flow during 2017 freshet on the Nicola River, downstream of Guichon Creek. .......... 77 Figure 3.3. Diel temperature maxima (°C) across all sites during 2020 and 2021 data collection periods. Blue data series indicate thermal refuges while red data series (open points) indicate adjacent Nicola River mainstem. .................................................................................................. 78 Figure 3.4. Boxplots of fork lengths of focal species captured during project sampling. ............ 79 Figure 3.5. Diel detection frequencies of fish at all sites across 2020 and 2021. Site 2S is represented in solid grey within the same panel as Site 2. Note that Site 2S is a single antenna reader that detected occupancy at that thermal refuge. ................................................................. 81 x Figure 3.6. Diel migration frequency by Site. Note that migration direction could not be interpreted for Site 2/Site 2/2S. ..................................................................................................... 82 Figure 3.7. Emigration (E), Immigration (I), and Detection (D) times associated with Sunrise (SR) and Sunset (SS) events at all sites. ....................................................................................... 84 Figure 3.8. Partial effects plots of individual smoother terms on thermal refuge occupancy with 95% confidence intervals. Row 1 displays partial effects on log-odds scale. Row 2 has been converted to probability scale and incorporates model intercept and uncertainty. y-axis labels show effective degrees of freedom of smoother term, indicating degree of wiggliness. .............. 86 Figure 3.9. Partial effects plot of thermal refuge occupancy displaying interaction between model terms Time and Mainstem Temperature. Points indicate discrete occupancy records across Time and Mainstem Temperature used in the model. ............................................................................ 87 Figure 3.10. Partial effects plots of individual smoother terms on DHM detections at Site 3 with 95% confidence intervals. y-axis labels show effective degrees of freedom of smoother term, indicating degree of wiggliness. ................................................................................................... 88 Figure 3.11. Partial effects plot of DHM detections at Site 3 displaying interaction between model terms Time and Mainstem Temperature. ............................................................................ 88 xi Acknowledgements I would first like to thank and acknowledge the Indigenous Peoples of the Nlaka'pamux and Syilx Nations, recognizing that my research activities took place on their unceded, ancestral, and traditional territories. I acknowledge that I reside as an uninvited guest here and am humbled that I have never been made to feel unwelcome. To paraphrase Grand Chief Percy Joe’s invitation to those not Indigenous to the Nicola Valley - I know I am welcome in your home but will try not to move the furniture around. Kʷukʷsc̓émxʷ and limləmt (thank you). To my co-supervisors, Dr. Tom Pypker and Dr. Mark Shrimpton, thank you for sharing this role so graciously and for fostering collaboration on what were diverse research interests. Thank you, Tom, for enduring this academic journey with me from master’s student to PhD candidate, and for helping me grow my appreciation for physical sciences, as most of my fish questions were actually hydrology questions. Thank you, Mark, for taking me on as a PhD student and for your advice, edits, conversations, and questions that were always kind and constructive. I appreciate both of you for your patience during long waits in communication (by me) punctuated by flurries of requests, writing, and edits. I had a great experience under both of your supervision. I would like to thank my committee members Dr. Eduardo Martins and Dr. David Hill. Thank you, Eduardo, for helping me get started in R and for valuable advice, code, and literature relating to my thermal refuge occupancy model. Thank you, David, for staying on my committee from master’s student to PhD, and particularly for your help with the remote sensing aspects of my research. Thank you also to my former committee members at Thompson Rivers University, Dr. Brian Heise and Dr. John Church, whose advice and assistance helped me in the initial years of this research. Thank you to my fellow graduate students for all of the great conversations and for the occasional late night working on group projects. Thanks to Dr. Garrett Whitworth for being a great partner in our early work on stream thermal imaging with drones, including flying them (even sometimes upside down) and for your efforts in processing thousands of thermal images and stitching what seemed like endless batches of orthomosaics. Thanks to Peter Grandia for helping me solve my database problems – you are truly an expert. Thank you to Nicola Valley Institute of Technology students Jerome Abbott and Jerrod Peterson who helped me with field data collection during their summer co-op terms. Thanks to Luc Turcotte for sharing ideas and for help in the field. Thanks to all of the members of the former Nicola Research and Technical Committee who showed interest in my project and provided feedback, specifically: David Lawrence, Richard Bailey, Dr. David Reid, Luke Warkentin, Dr. Jonathan Moore, Tracy Thomas, Mike Simpson, Brian Holmes, Leonna Antoine, Patrick Farmer, Christian St. Pierre and Dr. Richard McCleary. Thank you to Aaron Boone at Alpha Welding for fabricating numerous tools and equipment for my research that were not commercially available, and for donating all time and materials. Thanks to the following landowners for access through their properties: Robert Hack, Adam and Jennifer Van Leeuwen, and Chilliwack Cattle Sales Ltd. – Merritt Division (thanks, Hap!). xii I would like to thank my parents, Paul and Joan Willms, for their love and support throughout the years. To my kids, Johanna, Gretchen, Beck, and Peter, thank you for your inquisitive minds and for being patient when Dad was away or just distracted. Not least of all, I would like to thank my wife, Deborah, who so often put my ambitions first. Thank you for your support in so many tangible and intangible ways. Truly, this would not have been possible without your help. Funding External funding for this work was provided, in part, through the Government of Canada’s Habitat Stewardship Program as well as Sustainable Forestry Initiatives grants through the Fraser Basin Council. xiii Prologue Pacific salmon (Oncorhynchus spp.) and steelhead (O. mykiss) are cold-water species, having geographic distributions that include rivers and marine environments of the northern Pacific Ocean and proximal regions of the Arctic Ocean. At the southern extents of their range, stream temperatures can exceed critical thresholds for survival during the summer, affecting adult migration and spawning, as well as rearing for river-type populations where juveniles have a prolonged period of freshwater residency prior to smolt migration. Climate-driven increases in stream temperature are partly due to increases in one of its primary heat fluxes, air temperature (Caissie 2006), but are most strongly attributed to associated reductions in stream discharge during the summer (Poole and Berman 2001; Cristea and Burges 2010; Isaak et al. 2012; Woltemade and Hawkins 2016; Warkentin et al. 2022). Changes in stream temperature can drive adaptation and microevolutionary processes in salmon populations (Reed et al. 2011) but much concern exists as to how future stream temperature conditions will affect the biological and population-level responses of fish under various climate change scenarios (Yates et al. 2008; Keefer et al. 2018; Zhang et al. 2019; Reeder et al. 2021). Simulations of future climate conditions predict contraction of Pacific salmon and steelhead distributions in southern extents of their range, particularly in interior, snow-dominated watersheds (Mantua et al. 2010; Wade et al. 2013; Crozier et al. 2021). Streams of these watersheds rely on transient storage from snowmelt during the summer months and are most vulnerable to the effects of drought and low-flow associated heating during critical rearing periods for stream-type salmon and steelhead (Mantua et al. 2010). Salmon are poikilothermic, relying on movement and availability of suitable temperature environments, rather than internal physiology to regulate their body temperature. It has long been 1 observed that stream-dwelling salmonids differentially exploit discrete patches of cold water often associated with groundwater upwelling and tributary confluences - during periods of high ambient stream temperature conditions. It was only in the 2000s, however, that the first empirical evidence began to emerge and the associated effects of these types of habitats on fish behaviour, abundance, and density could be quantified and modelled (e.g., Ebersole et al. 2001, 2003b; Breau et al. 2007, 2011). Habitats where fish move to escape inhospitable ambient temperature conditions were often termed cool water refugia or thermal refugia, but more recently, Sullivan et al. (2021) more accurately defines these discrete, cold-water sites that are used by poikilotherms as thermal refuges. Since first introduced, the concept of thermal refuges for salmon in streams has gained popularity, coincident with research evidence and concern regarding climate-induced warming of stream thermal regimes. Where modelling of future climate and stream thermal conditions, as well as current empirical evidence, would limit the effectiveness of conservation options directed at vulnerable salmon populations (due to high stream temperatures being a limiting factor) numerous studies describe uncertainty in terms of the potential resilience provided by stream thermal refuges during temperature extremes (Poole and Berman 2001; Mantua et al. 2010; Woltemade and Hawkins 2016). In British Columbia’s Southern Interior region, semi-arid landscapes occur on the leeward slopes of major mountain ranges (e.g., Coast Mountains and Northern Cascades). Streams in these landscapes are generally considered temperature-sensitive in terms of their response to air temperature conditions, which are typically hot during the months of July and August. Consistent with modelling, current empirical evidence suggests that salmon streams in this region are experiencing lower productivity from reduced summer baseflows and higher temperatures during juvenile rearing, and that these observations coincide with climate change 2 and anthropogenic disturbance (Warkentin et al. 2022). Within their juvenile residency period in freshwater, survival and growth of Pacific salmon and steelhead is not simply a function of mainstem stream temperatures as many streams have been shown to be thermally heterogeneous, and juveniles of these species exploit this heterogeneity through diel horizontal migration (DHM) (Armstrong et al. 2013; Brewitt and Danner 2014; Brewitt et al. 2017). Studies by Armstrong et al. (2013), Brewitt and Danner (2014), and Brewitt et al. (2017) have provided insight into how juvenile Pacific salmon and steelhead balance trade-offs between, temperature, metabolism, and food availability through DHM. These trade-offs are complex and can be markedly different under diverse climates and stream thermal regimes. For instance, juvenile Coho Salmon (O. kisutch) at more northern latitudes make DHMs between cold-water feeding areas and warmer groundwater-influenced sites where digestion and assimilation of food was improved (Armstrong et al. 2013). Alternatively, at more southern latitudes, steelhead have been found to occupy the relatively cooler conditions found in tributary confluences but will feed primary in warmer mainstem habitats (Brewitt et al. 2017). These scenarios reflect trade-offs between feeding and metabolism between disparate thermal environments; however, juvenile salmonids also move out of mainstem habitats into thermal refuges in response to diel temperature fluctuations reaching critical thresholds (Breau et al. 2007, 2011; Brewitt and Danner 2014). Ecological complexities related to stream thermal refuges reflect the diversity in salmon species, subpopulations, life-history strategies, geographies, and climates that are unique at each scale of study. Few studies have paired physical processes at thermal refuges with biological processes at fine scales. My research describes processes in identifying thermal refuges in streams; characterizing the physical conditions, and interactions between, streamflow, hyporheic 3 exchange and temperature; and, characterizing behavioural responses of juvenile salmonids to environmental cues at thermal refuges. In Chapter 1, I describe novel techniques for identifying and characterizing thermal refuges in warm streams using low-cost remotely piloted aircraft system (RPAS)-based thermal infrared (TIR) imaging. As part of this work, I examined spatiotemporal variability in thermal refuges within my study reach before and after a major channel-forming flood event. My work in this chapter contributes to a growing body of research on fisheries applications of RPAS-based TIR imaging. It also highlights the importance of flood disturbance and functional floodplains in maintenance and recruitment of thermal refuges in streams. In Chapter 2, I selected a subset of identified thermal refuges from the work described in Chapter 1 and installed temperature loggers and shallow streambed piezometers to characterize in situ temperature and hyporheic exchange conditions over two summers. Thermal refuge temperature, mainstem river temperature, and groundwater upwelling at thermal refuges were modelled in response to relevant environmental covariates, including: mainstem river temperature, air temperature, and stream discharge. Modelled responses were highly variable by site with notable differences between mainstem and off-channel thermal refuges. I discussed the importance of my results as related anthropomorphic changes to river corridors, manipulation of flows from dam releases, and uncertainty around future flow and temperature conditions due to climate change. In Chapter 3, I studied patterns of DHM in juvenile salmonids at thermal refuges that were identified and characterized in Chapters 1 and 2, respectively. Over two field seasons, I logged individual fish movements during summer and early fall between mainstem and thermal refuge habitats using Passive Integrated Transponder (PIT) tag technology. DHMs and juvenile 4 salmon occupancy at thermal refuges were modelled in response to mainstem and thermal refuge temperatures as well as photoperiod. My results bring new insights into how and when fish use thermal refuges and have important implications in managing for ecosystem resilience considering uncertainty related to current and future changes to stream thermal regimes. 5 CHAPTER 1 Identification of thermal refuges in streams using remotely piloted aircraft system-based thermal infrared imaging 6 Abstract In this study, I used remotely piloted aircraft system (RPAS)-based thermal infrared (TIR) imaging to study fine-scale thermal heterogeneity in a temperature-sensitive salmon stream in the Southern Interior region of British Columbia. I was able to identify and classify cold-water patches, in a generally warm stream, that may provide important thermal refuges for salmonids during periods of high summer stream temperatures. Repeat imaging of my study reach following a major flood event in 2017 found that the abundance of cold alcoves, a type of thermal refuge habitat, increased significantly. These results highlight the importance of maintaining the fluvial processes that create new habitat, in-keeping with the concept of Shifting Habitat Mosaics. As part of this work, I also provide some best practices for rapid, RPAS-based TIR surveys of stream thermal refuge habitats – expecting that this type of survey will become an increasingly common addition to fish habitat inventories. 7 Introduction The use of aerial thermal infrared (TIR) imaging from small, remotely piloted aircraft systems (RPAS) is becoming increasingly accessible to researchers and practitioners. Applications of this technology vary from detection of underground fires (Burke et al. 2019) to identifying differences in tree genotypes through thermal variations exhibited during drought stress (Ludovisi et al. 2017). Stream temperature applications of RPAS-based TIR imaging have been steadily increasing since first presented by Jensen et al. (2012). TIR imaging allows for rapid visualization and increasingly effective measurement of thermal heterogeneity in streams at multiple scales. TIR images are collected using a thermal imager that detects emittance of TIR radiation from the surface of objects. Traditionally, TIR remote sensing has been conducted from piloted aircraft (e.g., helicopters and planes), with the earliest fisheries applications of TIR technology emerging in the late 1990s and early 2000s from Torgersen et al. (1999, 2001). However, recent advances in technology have allowed for production of TIR sensors that are much smaller in size, lighter, more affordable and are designed for use with RPASs (Hill et al. 2019). These imagers are generally uncalibrated, exhibiting relative temperature values that are only accurate in terms of their differences to adjacent pixels. Recently, however, work has been done to develop methods to improve radiometric-calibration for situations where estimates of kinetic temperature are needed (Jensen et al. 2014; Kelly et al. 2019). Although uncalibrated, these systems offer advantages to researchers and practitioners in terms of accessibility and ease of use in analysing patterns of fine-scale thermal heterogeneity in streams. In British Columbia, Canada and the Pacific Northwest, USA, some streams of temperate, semi-arid regions lack advection from glacial melt or have limited summer baseflows. In these 8 streams, ambient stream temperature conditions are highly sensitive to atmospheric temperature conditions (Moore et al. 2013), and often exceed upper thresholds for salmonid growth and survival (Kosakoski and Hamilton 1982; Walthers and Nener 1997; Ebersole et al. 2001; McCulloch et al. 2001; Richter and Kolmes 2005); however, many streams have been shown to be spatially heterogeneous in their thermal patterns, even at fine scales (Ebersole et al. 2003a and 2003b; Dugdale et al. 2013, 2015 and 2019; Casas-Mulet 2020). This thermal heterogeneity can be the result of the advective cooling of mainstems from tributaries (Torgersen et al. 2012; Brewitt et al. 2017), groundwater-fed off-channel habitats (e.g., cold alcoves/snyes and sidechannels) (Ebersole 2003b), discrete patches of groundwater exfiltration (Ebersole et al. 2003b; Kurylyk et al. 2015a) as well as diffuse groundwater-surface water interactions (Burkholder et al. 2008). Diel variability in groundwater temperatures is strongly attenuated by its interaction with the relatively stable temperature of the earth. Because of this, stream temperatures tend to be closest to groundwater temperatures when near their source (Caissie 2006) (e.g., patches of exfiltration or in spring-fed channels). The influence of groundwater on stream temperature variability tends to decrease from headwaters to higher order watercourses but remains an important heat flux in larger streams (Guenther et al. 2012; Caissie 2006). Precipitation is infrequent during summer months in semi-arid climates, so groundwater is critical in maintaining environmental flows for aquatic ecosystems and for anthropogenic needs (McCallum et al. 2013). Besides the advective cooling effects of groundwater on larger streams, discrete sources of groundwater in streams create thermal heterogeneity which can be differentially exploited by fish during suboptimal ambient stream temperature conditions (Ebersole et al. 2003a; Armstrong et al. 2013). In streams where temperatures reach physiological limits for fish growth and survival, these habitats have been termed thermal refuges (Sullivan et al. 2021). 9 For the purposes of this study, I have limited the definition of thermal refuges to proximal and accessible aquatic habitats that are ≥ 2°C cooler than average ambient stream temperatures. The temperature limits of this definition are in-keeping with those used in other literature (Torgersen et al. 2012; Fullerton et al. 2018), but some definitions have used differences of only 0.5°C to constitute thermal refuges (Dugdale et al. 2013) based on real-world thermal selectivity by salmonids. In addition, others have used temperature differential of > 3°C to define thermal refuges (Ebersole et al. 2003b). Regardless of chosen definition, it is important to note that behavioural thermoregulation in salmonids is not binary but represents complex interactions between organism and environment whereby survival must be ensured, and metabolism optimized. Sullivan et al. (2021) refer to thermal refuges as being a “melting pot”-type concept that incorporates hydrology, biology, and ecology in defining areas that provide some spatially discrete thermal relief for poikilotherms from adverse temperature conditions of the surrounding habitat. Much of the research regarding thermal refuges focuses, not surprisingly, on temperature; however, Kurylyk et al. (2015a) notes that trade-offs exist at thermal refuges when non or indirect temperature-related habitat factors (e.g., cover from predation hazards and anoxic groundwater conditions) reduce their suitability for target species, and so need to be validated as biologically significant as well. Adding to the thermal, hydrological, and biological complexity of thermal refuges, evidence suggests that thermal refuges within a stream system can be dynamic, with some persisting through time (e.g., cold tributary confluences), yet others being either ephemeral or strongly linked to specific flow conditions and precipitation trends (Dugdale et al. 2013). Conservation of thermal refuges is generally based on spatially static features that once identified, can be preserved or protected by various means (Casas-Mulet et al. 2020; Kurylyk et al. 2015a). 10 Temporal and spatial dynamics of thermal refuge and groundwater upwelling have been studied in response to changing stream flow conditions, and the associated effects to stream stage and recharge of alluvial aquifers, within static channel alignments (Dugdale et al. 2013; Käser et al. 2009); however, channel alignments and habitat features in alluvial rivers of unconfined floodplains can change dramatically between years because of larger-scale flood pulses. These events strongly influence the composition of habitat patches and biotic communities (Kleindl et al. 2015; Hauer et al. 2003) and drive productivity and resiliency in aquatic ecosystems (Brennan et al. 2019). Research into the effects of large-scale flood events on the abundance, type, and distribution of stream thermal refuges is lacking. The research goals of my study were to assess the efficacy of conducting rapid, low-cost, RPAS-based TIR surveys in identifying potential thermal refuges in a temperature sensitive salmon stream; to examine changes in abundance and type of thermal refuges following a major channel-forming flood event; to analyze patterns of dispersion in thermal refuges in my study reach between successive years; and, to provide guidance on best practices for conducting rapid surveys of thermal heterogeneity in streams using uncalibrated radiometric imaging techniques, based on my own experience and similar studies. Methods Study Location In the Southern Interior region of British Columbia, Canada, the Nicola River and its tributaries flow through a diverse landscape from montane and subalpine ecosystems to open grasslands and sagebrush at lower elevations. The Nicola Watershed has an area of 7,211 km 2 and is climatically diverse in terms of temperature and precipitation regimes. The Nicola River flows generally east to west, draining the drier, Interior Plateau to the east and the wetter, coastal- 11 influenced Cascades Mountains to the west – eventually joining the Thompson River at the community of Spences Bridge, British Columbia. Precipitation regimes range from only 200 to 300 mm annually in valley bottoms to potentially over 2000 mm in some of the high-elevation, coastal-influenced ecosystems in southwest portions of the watershed (Lloyd et al. 1990). Precipitation regimes in eastern portions of the watershed are drier because of the rain shadow effect produced by the Coast Mountains to the west. These climatic differences within the watershed produce variability in hydrometeorological conditions, fish species, population assemblages, and life-history strategies between sub-catchments (Rosenau and Angelo 2003). The Nicola River’s annual hydrograph is typical of snow-dominated regimes, having peak flows in late-April through May followed by a steady decline towards base-flow conditions in late-August and September. Fall flooding is not typical in the Nicola Watershed, but in recent decades, the effects of climate change have been exhibited by more precipitation occurring as rainfall, resulting in higher flows in late-October and November. Intra-annual shifts towards higher amounts of rain vs. snow across many parts of North America (Ryberg et al. 2016) can disrupt aquatic ecosystems and reduce mean annual discharge in streams (Berghuijs et al. 2014). In the Nicola River watershed, the two primary energy fluxes to streams, direct solar radiation and convective heating (Caissie 2006), are exacerbated by degraded riparian ecosystems (Rayne et al. 2008) and atmospheric conditions that are known to exceed 40°C (Walthers and Nener 1997). Subsequently, streams regularly experiences high water temperatures between 25 and 30°C during the summer months (Kosakoski and Hamilton 1982; Walthers and Nener 1997), which exceed established growth and survival thresholds for juvenile Chinook Salmon (Oncorhynchus tshawytscha), Coho Salmon (O. kisutch) and steelhead (O. mykiss) (McCulloch et al. 2001; Richter and Kolmes 2005) that rear in the system, but also for returning 12 adults during migration and spawning. Populations of all three of these species in my study location are of conservation concern and have been in decline in recent decades (Mathews et al. 2007; Dobson et al. 2020). Within the Nicola River, I chose a 22.4 km study reach (Figure 1.1) that is temperaturesensitive and known to have high-value spawning and rearing habitat for salmon and steelhead. The extents of my study reach were defined by its confluence with the Coldwater River at the upstream end and Spius Creek at the downstream end (Figure 1.1). This reach has a low-gradient (0.3%), sinuous, riffle-pool-run morphology and has been reported to be used by as much as 75% of the Nicola River’s late-run Chinook Salmon for spawning (Kosakoski and Hamilton 1982). It is also strongly influenced by groundwater, as seen by extensive networks of wetlands and springbrooks associated with floodplain depressions. The Coldwater River is an important spawning stream for early-run Chinook and, along with Spius Creek (and its tributaries), provide the bulk of spawning habitat for Coho and steelhead in the Nicola Watershed. Individuals of all three species’ populations typically exhibit a period of freshwater residency as juveniles (1-2 years) before migration to the ocean as smolts. Chinook Salmon and Coho Salmon juveniles from the Coldwater River have been found to migrate downstream to the Nicola River mainstem to rear prior to smolt transformation (Shrimpton et al. 2014). Considering the potential habitat use of this reach, its extensive groundwater influence, and the high ambient stream temperatures experienced during the summer, it appeared to be an excellent model system for my study. 13 Figure 1.1. 22.4 km study reach (denoted by thickened grey line) located on the Nicola River between its confluences with Coldwater River (upstream) and Spius Creek (downstream), near the city of Merritt, British Columbia, Canada. Equipment TIR data collection for this project was achieved with small, RPAS-mounted TIR imagers that use uncooled microbolometers to produce radiometric images that display relative/uncalibrated temperatures. The TIR sensor used for 2016 and 2017 imaging was a thermal infrared imager (336 x 256 resolution, 9 mm lens and a 30Hz refresh rate) (ZenMuse XT, DJI, Shenzhen, Nanshan District, China) attached to a professional RPAS (Inspire 1, DJI, Shenzhen, Nanshan District, China). Zenmuse XT imagers are detachable from the RPAS and have a gimbal to reduce vibration and maintain image geometry during flight. 14 Zenmuse XT imagers can store images as simple JPEGs, radiometric JPEGs or TIFF files. All image storage formats include location data, however, it is important to note that the accuracy of the GPS receiver on the UAV is considered ‘recreation grade’ (i.e., does not collect the required carrier data for differential correction) and was expected to have an accuracy of ± 10 m. Field Methods Two iterations of imaging were conducted across the entire study reach in September of 2016 and August of 2017. Initially, just one TIR survey was planned, for 2016; however, the freshet flows of 2017 produced extensive flooding across the study reach (and the Nicola Watershed) and resulted in shifts in channel alignment, sediment erosion and aggradation, and changes in habitat complexity and features. This provided me with an opportunity to assess changes to the distribution and composition of potential thermal refuges following a major hydrologic event. The timing of data collection was intended to coincide with summer low-flow conditions and high stream temperatures, as described by Torgersen et al. (2001). The rationale for this is to maximize the effects of cool groundwater and tributary input on stream temperatures and to produce the strongest temperature signature visible in the TIR imagery. In applying TIR surveys to identification of summer thermal refuges, this timing also coincides with expected peak use by salmonids (Breau et al. 2007, 2011). For the Nicola Watershed, maximum stream temperatures typically occur at the beginning of August, however, base-flows are not typically reached until late August or early September. Stream discharge has not been shown to produce strong effects on hyporheic exchange processes (Wondzell 2006; Marzadri et al. 2014) but lower stream flows result in less 15 dilution of upwelling groundwater, which improves its detectability in TIR images. Daily stream temperature maxima for the Nicola River were not typically reached until about 16:00, but potentially adverse temperatures for salmonids (i.e., >20°C) were common by 12:00. Solar noon occurs around 13:00 in August, which is the optimum image collection time, specific to optimizing the angle of the sun and reducing reflectance of solar radiation. As a result, TIR surveys were generally planned for times between 11:00 and 16:00, which is consistent with similar studies (Dugdale et al. 2015; Casas-Mulet et al. 2020). One issue that consistently proved to be problematic during my surveys was the heat-generated atmospheric winds that are common in the Nicola Watershed and in other semi-arid regions with hot climates. These warm winds cause several issues during TIR image collection, including: increased roughness on the surface of the water resulting in differential emissivity, as described by Torgersen et al. (2001); distortion caused by convective heating and cooling across the focal plane of the imager, as described by Kelly et al. (2019); and, RPAS instability and subsequent reduced flight times as a function of increased power requirements. RPAS flight operations were conducted by floating sections of the study reach in a drift boat with the RPAS operator on the bow, operating the RPAS and collecting data almost continuously. This type of continuous flight operation became challenging with respect to battery supply but allowed me to complete the TIR imaging of the 22.4 km study reach in only two sessions. Flight times, based on battery life, varied in response to wind conditions (i.e., higher winds increased the power required for RPASs to maintain their position during flight) but were typically between 15 – 20 minutes. Battery supply can be a major challenge to larger-scale flight operations. I was able to plan for this by bringing up to four sets of batteries in the field and by charging two sets of batteries at a time from a generator I loaded and secured in the drift boat. 16 Images were collected at a targeted altitude of 60 m with the imager in nadir position. The targeted altitude was determined based on the lowest required height above ground to capture the average width of the channel in one frame, to promote high image resolution. The RPAS was piloted manually using DJI’s GO app, with images being collected every five seconds. In 2020, I switched to DJIs XT app, which is designed for use with DJI’s XT imagers. The RPAS was flown down the stream centreline when the field of view of the imager was able to capture both streambanks in a single image. When the river was too wide to capture in a single image, the RPAS was piloted towards channel margins and off-channel habitats to capture the remaining segments of stream. Data Processing Post-processing and analysis of collected images was completed using FLIR Tools (Version 5.11.16357.2007). An arctic color palette was applied to the images, and they were assessed individually for areas that displayed cool-water signatures. In some cases, displayed temperature ranges were scaled to a narrower range to produce higher contrast symbology for easier identification of fine-scale temperature anomalies. These images were further analyzed by determining the uncalibrated water temperature at points representative of ambient water temperature conditions and comparing these to relative temperatures at identified thermal anomalies within the same image. In this way, the difference in temperature between the identified cool-water zone and the average background stream temperature could be determined (Figure 1.2). 17 Figure 1.2. Example of thermal infrared image showing differences in relative temperature between river mainstem and identified thermal refuge habitat. Sp1 to Sp4 indicate spot measurements of relative temperature and corresponding locations. Where in situ measurements of water temperature were recorded at thermal refuges or at background sites, these measurements could be used to correct temperature values in images and better approximate absolute temperature values; however, it is important to note that because I did not collect extensive in situ temperature measurements during this work, no further radiometric calibration was conducted – other than that which is performed as part of the imager’s own internal proprietary radiometric calibration. Data processing was completed using the previously described “black box” technologies that are readily available to consumers. Images that had visible cool-water zones that were ≥ 2°C cooler than ambient conditions were identified, classified according to seven standardized thermal refuge classes (Table 1.1) (Torgersen et al. 2012; Dugdale et al. 2013; Dugdale et al. 2015) and locations were added to a Geographic Information System (GIS) using ArcMap (Version 10.6) (ESRI, Redlands, California, USA) to produce a map. 18 Table 1.1. Descriptions of thermal refuge habitat classes used in data analysis (Modified from Dugdale et al. 2013, 2015). Thermal Refuge Class Description Cold side-channel Cool seepage flow from ephemeral channels adjacent to the mainstem that are expected to be annually connected to the river at their upstream end. Wall-base channel Cool water discharge from channels originating along the base of valley walls, and from terraces that are above the height of the floodplain. Tributary confluence plume Plume of cold water from tributary discharge into mainstem habitats. Springbrook Surface water discharge generated within the adjacent floodplain from springs and wetlands. Lateral seep Discharge of groundwater along channel margins under positive hydraulic gradients. May be diffuse or discrete (when preferential flow paths are present). Hyporheic upwelling Cool water zones immediately downstream of riffles, bars and meanders resulting from exfiltration of macro-habitat scale hyporheic water. Cold alcove Cool water habitats that are connected to the mainstem at their downstream end only - usually associated with lateral channel migration and abandoned/relic channels. With my sites in a GIS, density of thermal refuges was calculated between years, based on the average number of thermal refuges per river kilometer (stream centreline). To estimate the degree of clustering, or uniformity, in spacing of thermal refuges within the study reach, I used an index of dispersion based on counts of thermal refuges, with individual river kilometers being the sample unit. I also performed an Optimized Hotspot Analysis in GIS using the Getis-Ord Gi* statistic (Getis and Ord 1992; Ord and Getis 1995) to confirm the dispersion patterns of my perriver-kilometer count data. The analysis was limited to a 200 m buffer of the study reach stream centreline which captured all identified within-channel and off-channel thermal refuges. Results Density and composition of thermal refuge habitats In total, I collected over 8000 thermal images and identified 64 potential thermal refuges between 2016 and 2017. The density of thermal refuges in the study reach increased between 2016 and 2017 from 1.23 km-1 (27 total) to 1.68 km-1 (37 total), respectively. Total thermal 19 refuges by class are provided in Figure 1.3. Using McNemar’s test of symmetry (c2, p < 0.05, df = 1), significant shifts in thermal refuge classes were observed only for cold alcoves, which increased from two (2016) to 16 (2017). Counts of all thermal refuge classes per river km indicate random dispersion, according to their dispersion indices (s2/x̄) (1.00 in 2016 and 1.32 in 2017, n=22) within the study reach, with aggregation increasing slightly in 2017. Chi-square values, the product of dispersion indices and df for 2016 and 2017 thermal refuge counts, fall within the upper and lower 95% confidence interval for random dispersion, indicating no significant deviation of the variance from the mean in number of thermal refuges per river km, as would be seen with clumped or uniform dispersion patterns. Random dispersion was also confirmed using an Optimized Hotspot Analysis in ArcGIS based on the more aggregated 2017 thermal refuges. No significant (95% confidence) hot or cold spots were identified. Figure 1.3. Counts of identified thermal refuge habitats by class between 2016 and 2017 TIR surveys. 20 Composition of identified thermal refuges by class varied only slightly in all classes but cold alcoves between 2016 and 2017. In 2016, lateral seeps were the dominant class, comprising 56% of all thermal refuges identified. Springbrooks were the subdominant class in 2016, comprising 30% of identified thermal refuges. Counts of identified thermal refuges in 2017 were similar to 2016 with the exception of cold alcoves which increased eightfold, becoming the dominant class and comprising 43% of all identified thermal refuges. Subdominant categories in 2017 included lateral seeps and springbrooks (30% and 24% respectively). It is important to note that changes in identified thermal refuges may be due to any combination of changes to physical habitat and hydrometeorological conditions at the time of the survey. For instance, one tributary thermal refuge was identified in 2016 (Guichon Creek) that had a confluence plume temperature 2.3°C cooler than the mainstem of the Nicola River. In 2017, no tributary refuge habitats were identified as Guichon Creek’s temperature was only 1.5°C cooler than the Nicola River at the time of the survey, therefore not meeting my definition of thermal refuges. This anomaly does not represent changes in physical habitat but factors in diel and seasonal temperature shifts. The increased presence of cold alcoves in 2017, on the other hand, were generally attributed in the field to the substantial channel shifts which occurred during the spring flooding of that year. I used available peak flow records from 1912-2020 (n=71) for the Nicola River at Spences Bridge (Water Survey of Canada 2021) and found that the 2017 spring flood event on the Nicola River represented a 1:36 year recurrence interval that resulted in overbank flow, lateral channel migration and shifts in sediment aggradation and erosion across the entire study reach. 21 Discussion Effectiveness of methods in identifying thermal refuges The combination of DJI’s professional RPASs and XT imagers provided a stable platform for rapid thermal imaging of my study reach and identification of potential thermal refuges. I found FLIR Tools to be an effective software package that allowed for rapid visualization of thermal heterogeneity and anomalies within individual images. The ability to select different colour palettes (symbology) and scale displayed temperature ranges allowed me to maximize the visual differences within selected temperature ranges, thus improving my ability to identify thermal anomalies in streams. Although not critical for simply identifying potential thermal refuges, the difference function in FLIR Tools provided a simple mechanism for referencing absolute temperatures (from in situ measurements) in uncalibrated TIR imagery (Figure 1.4). These corrections helped approximate kinetic temperatures, when needed; however, similar studies have found these corrections to be unreliable due to the complexity of interactions between environmental conditions and RPAS flight characteristics (Dugdale et al. 2019). 22 Figure 1.4. Image of identified thermal refuge habitat with temperature difference value (-5.4˚C) based on in situ measurement of mainstem habitat temperature at point "Sp2" and thermal refuge habitat temperature at point “Sp1”. Absolute temperatures at Nicola River mainstem and thermal refuge habitat and would be estimated at 19.2˚C and 14.1˚C, respectively. During my analysis, effective interpretation of TIR images in identifying thermal heterogeneity in streams required an understanding of the effects produced across the focal plane of the sensor and an understanding of the effects of distortion within images. Many of the apparent thermal anomalies within images were due to differences in emissivity between objects. For instance, north facing stream banks that were shaded produced temperature signatures in images that could easily be mistaken for groundwater seepage along a channel margin (Figure 1.5). Differences in surface roughness between stream macrohabitat types (e.g., runs, riffles and pools) also resulted in different rates of emissivity and apparent variation in temperature in TIR images (Torgersen et al. 2001). 23 Figure 1.5. Example of shaded, north-facing stream bank has the potential to be misidentified as thermal refuge – specifically, lateral seepage. Conspicuous cool-water zones towards the edges of images were also observed occasionally that were due not to actual thermal anomalies, but to vignetting. Vignetting is a thermal artifact that results in reduced brightness towards the edges of images. This effect was rarely observed in my radiometric images in FLIR Tools, however Alekseychik et al. (2021) observed vignetting effects causing temperature differences between focal centre and margins exceeding 2°C using a similar Zenmuse XT2 thermal imager. In similar studies to mine, solar reflections have also been observed to cause warming gradients across images as much as 4°C, even with the imager in nadir (Dugdale et al. 2019), and convective heating and cooling across the focal plane of the imager can cause interpretation issues between images (Kelly et al. 2019). In the absence of accompanying RGB imagery, thermal anomalies associated with water were best identified by the presence of a thermal plume – associated with the downstream mixing effects of cool water with warm water (Figure 1.6). Torgerson et al. (2012) and Dugdale et al. (2013) also provide excellent examples of thermal plumes associated with the previously noted thermal refuge classes which were referenced during image analysis. 24 Figure 1.6. Evidence of downstream thermal plume can help differentiate potential thermal refuge habitats from thermal artifacts in TIR images. Image shows mixing of cool water from a springbrook thermal refuge habitat. Suggested Best Practices for collection of TIR images and identification of thermal refuge habitats Based on my results in the field and during image analysis, I propose the following bestpractices for rapid collection of TIR imagery in identifying potential thermal refuges in streams: (1) Collect images in a static position (i.e., hovering) with the imager in nadir position and attached to a gimbal, rather than hard-mounted to the RPAS. This optimizes the angle at which emittance from objects are detected by the sensor array and helps to reduce image distortion associated with movement; (2) Restrict timing of image collection to low-flow periods and during ambient conditions which maximize the difference between groundwater and surface water temperatures (Torgerson et al. 2001; Dugdale 2013, 2015). This applies to both diel and seasonal variation in temperature. Prior knowledge of diel and seasonal temperature and flow patterns will allow for more informed decisions in TIR RPAS mission planning; (3) In regions where heat-generated atmospheric winds are common, consideration should be given to planning 25 thermal image collection during the morning, prior to the generation of afternoon winds, and ceasing data collection when winds become too strong. This timing deviates from those used for helicopter-based TIR surveys (Torgerson et al. 2001; Dugdale 2013, 2015), but helicopters are understandably more resistant to the adverse effects of wind during flight operations than RPASs. I suggest limiting RPAS-based TIR surveys to wind speeds ≤ 10 km/hr (about a “2” (light breeze) according to the Beaufort scale); (4) Maintain flight altitudes that allow for imaging of both stream banks in a single image (Dugdale et al. 2015). This altitude will depend on the width of the stream but also the focal length of the imager. (5) Ensure ample overlap between successive images. Similar work suggests as much as 60% (Torgerson et al. 2001; Dugdale 2013, 2015). This will help prevent thermal refuges not being detected and will also allow for analysis of cool water anomalies from different image angles. Also, where creation of orthomosaics is needed, generous overlap between images will allow photogrammetric software sufficient key points between images. Pix4D recommends 90% image overlap (front and sides) for orthomosaic production. I attempted to produce TIR orthomosaics from my 2016 flights with limited success. As I generally maintained a single flight path down the stream centreline, the image overlap was often not sufficient. Orthomosaic production was not the focus of this work and was not planned for in either the 2016 or 2017 flights. Production of orthomosaics from thermal images is inherently difficult due to being limited to one band of data, versus three in RGB imagery, reducing the key points available for generation of thermal reflectance maps by program algorithms. Generation of orthomosaics also requires consistent temperature scaling between images or must be calibrated for absolute temperature to avoid conspicuous bands of temperature changes between stitched images. Calibration to absolute temperature involves substantially more survey effort, requiring the use of ground control points (GCPs) and collection of in situ 26 temperatures which were not in-keeping with my rapid-survey methodology. When collecting TIR images in the field for orthomosaicking, one must also consider the longitudinal temperature drift that will occur within and between RPAS missions from changes in stream temperatures and atmospheric conditions (Casas-Mulet et al. 2020; Dugdale et al. 2019). Although difficult, substantial work has been contributed by others to improve orthomosaic production from TIR images which provide important layers for use in GIS and other applications (Casas-Mulet et al. 2020; Dugdale et al. 2019; Kelly et al. 2019). Spatio-temporal dynamics of thermal refuge habitats Rivers are dynamic ecosystems. The presence, form and abundance of thermal refuges are the result of landscape-level and hydrometeorological processes and have been shown to change through time (Dugdale et al. 2015). Although I did not attempt to link changes in thermal refuges between 2016 and 2017 to hydrometeorological conditions, the significant increase observed in cold alcoves is likely attributable to flood effects observed in 2017. Cold alcoves, unlike lateral seeps and hyporheic upwelling, are inextricably tied to a channel’s ability to avulse and move laterally across its floodplain. The resulting relic channel remains connected to the main channel at the downstream end and continues to be fed by cool groundwater. The concept of Shifting Habitat Mosaics (SHM), first introduced in the early 2000s (Hauer et al. 2003; Hauer and Lorang 2004), recognizes the importance of river dynamics and disturbance in maintaining essential habitats in streams and rivers. The idea of river connectivity, in which longitudinal connectivity had long been the primary focus, was extended to include lateral connectivity to floodplains and recognized that “cut and fill” processes in streams are integral to the retirement of discrete habitats and the formation of new ones. Research into the effects of artificial confinement of salmon streams has found that the fluvial processes that create 27 habitat become constrained, resulting in fewer refuges, including alcoves and side-channels (Blanton and Marcus 2013). At the watershed scale, SHM also become an important component of population resilience with respect to climate change, environmental stochasticity, and human-induced changes to habitat. Although salmon exhibit strong fine-scale natal site fidelity (Dittmann and Quinn 1996; Turcotte and Shrimpton 2020), interannual productivity shifts spatially across watersheds, providing an important buffering effect to populations through time (Brennan et al. 2019). In the Nicola Watershed, specifically, plasticity in salmon homing behaviour (Turcotte and Shrimpton 2020) and in habitat use during their freshwater life-history as juveniles (Shrimpton et al. 2014) have demonstrated the importance of protecting habitat and maintaining natural fluvial processes across entire watersheds. My results help to highlight the importance of keeping rivers connected to their floodplains but may also provide process-based insight into the role floods play in habitat maintenance and recruitment for salmonids. Role of RPAS-based TIR in conservation of thermal refuge habitats in streams Presence of thermal heterogeneity in the Nicola River is critical to providing adequate resources and the required diversity in environmental conditions to support the survival, growth, and reproduction of its salmonids. Juvenile salmon have been shown to use a diversity of thermal habitats for survival during periods of hot stream temperatures (Breau et al. 2007; Armstrong and Schindler 2013) and to optimize feeding and metabolism (Armstrong et al. 2013; Brewitt et al. 2017). The importance of thermal refuges in temperature-sensitive salmon streams is partly made apparent by the high densities of salmonids found in them during periods of high ambient stream temperatures (Breau et al. 2007, 2011; Brewitt et al. 2017). Additionally, abundance of salmonids in western streams has also been shown to respond positively to increased abundance of thermal 28 refuges (Ebersole et al. 2001, 2003a). Therefore, identification of potential thermal refuges across riverscapes is critical. As stream temperatures continue to increase with climate change (Isaak et al. 2012; Zhang et al. 2019), it is becoming ever more important to understand the extent and connectedness of thermal refuges; we need to conserve or enhance these habitats, and the fluvial processes that create them, to help support healthy populations of Pacific salmon and steelhead. I expect that rapid, RPAS-based TIR imaging will become increasingly common practice in fish habitat surveys of temperature-sensitive streams in the future. 29 CHAPTER 2 Vertical hydraulic gradient and temperature dynamics at stream thermal refuges under changing flow and atmospheric conditions 30 Abstract Alluvial rivers are thermally heterogeneous and involve complex interactions between the streambed, riparian area, flow, and atmosphere. Geomorphic features, flow conditions, and differences in streambed hydraulic conductivity create a mosaic of hyporheic exchange pathways which ultimately produce discrete patches of cooler water in warm streams during the summer. These anomalies, called thermal refuges, provide critical, yet limited, habitats in temperaturesensitive streams for cool-water species of fish, like salmon and trout. The objectives of my study were to characterize in situ thermal conditions and groundwater upwelling at discrete thermal refuges, or cold-water patches, as well as to examine changes in response to mainstem stream temperature, air temperature, and stream discharge in a river located in the semi-arid Southern Interior of British Columbia, Canada. I deployed temperature loggers, pressure transducers, and shallow streambed piezometers and found that mainstem stream temperatures were highly sensitive to changes in air temperature and, to a lesser degree, displayed sensitivity to changes in stream discharge. Associated mainstem thermal refuges were highly sensitive to changes in air temperature, while off-channel thermal refuges and sites exhibiting stronger groundwater upwelling were more resistant to changes in air temperature. The effects of stream discharge on hyporheic exchange processes, namely vertical hydraulic gradient, were variable, but at some thermal refuges I observed inverse changes in vertical hydraulic gradient in response to flow pulses from dam releases and atmospheric events. This study contributes to the collective understanding of temperature and groundwater conditions at thermal refuges in alluvial streams of semi-arid environments and provides important considerations regarding their conservation and management. 31 Introduction The expression of groundwater exfiltration in streams is recognized as a critical habitat component in freshwater thermal ecology and thermal refuges (Power et al. 1999; Kurylyk et al. 2014; Hitt et al. 2017; Ebersole et al. 2020), specifically when considering cool-water species such as Pacific salmon (Oncorhynchus spp.) and steelhead (O. mykiss) (Ebersole et al. 2003b). Groundwater provides discrete habitats with favorable thermal conditions for incubation (Curry 1995) and for rearing (Ebersole et al. 2003b; Armstrong and Schindler 2013), for example, but is also an important heat exchange flux in maintaining cool mainstem temperatures (Caissie 2006; Burkholder et al. 2008; Kurylyk et al. 2015a, 2015b) and adequate summer baseflows (Briggs et al. 2022; Murray et al. 2023; Singha and Navarre-Sitchler 2022). The maintenance of adequate stream flows and favorable temperatures during the summer are key components of productive freshwater habitat for salmonids, especially for those species and populations that have a prolonged freshwater residency period as juveniles. Among other things, stream flow directly affects the amount of physical habitat a stream provides, and functions to provide access to a stream’s habitat during migration events. In addition to this, the amount and timing of flow in a stream also affects its water temperatures (Poff et al. 1997; Woltemade and Hawkins 2016; Poff 2018), which has important implications for fish behaviour and physiology (Brett 1952; McCullough et al. 2001; Richter and Kolmes 2005). Climate change forecasts have been used to predict effects on stream temperatures and have found that degradation of stream thermal habitats is highest in streams with the lowest baseflows (Carlson et al. 2017). In the semi-arid, interior regions of the Columbia Basin, steelhead are projected to experience extreme low-flow conditions, and subsequently high temperatures, due to lower summer baseflows that will occur over longer durations (Wade et al. 2013). These semi-arid and 32 typically snow-dominated watersheds are experiencing shifts in their precipitation regimes and are becoming more rain-dominated, contributing to their propensity towards increasingly lower flows and high stream temperatures during the summer (Mantua et al. 2009). The risks to Pacific salmon and steelhead in these watersheds are high, however, projections of vulnerability do not account for thermal refuges created by groundwater inputs (Wade et al. 2013). Streams are thermally heterogeneous environments involving complex, three-dimensional patterns of temperature that undergo annual and diel changes in response to numerous heat fluxes. Heat fluxes applied at the surface of streams, namely short-wave solar radiation, long-wave radiation, evaporative heat loss, and convective heat exchange represent atmospheric conditions and are among the most important factors affecting stream temperature. Of these atmospheric conditions, direct heating as a result of incoming solar radiation is the most significant (Caissie 2006; Hebert et al. 2011). A key metric for grouped atmospheric heat fluxes, and one that is commonly used in modelling stream temperatures, is air temperature. Air temperature is generally understood to be the dominant factor affecting mainstem temperature regimes in higher-order streams (Caissie 2006; Isaak et al. 2012; Woltemade and Hawkins 2016), with increased effects in larger streams as one moves downstream from low-order, groundwater dominated headwater streams (Hebert et al. 2011). Air temperature is less responsible for patterns of thermal heterogeneity in streams, except to increase disparity between mainstem stream temperatures and sources of temperature anomalies (e.g., discrete groundwater upwelling). These temperature anomalies can provide important thermal refuges for aquatic organisms living in streams and are the result of numerous heat exchange processes. Examples include: advective mixing of surface water sources and changes in stream discharge (Woltemade and Hawkins 2016; Baker et al. 2018); exposure to solar radiation (Bray et al. 2017), or conversely, stream 33 shading (David et al. 2018; Dugdale et al. 2018; Cunningham et al. 2023); and conductive and advective hyporheic exchange processes (Burkholder et al. 2008; Caissie and Luce 2017; Surfleet and Louen 2018). Hyporheic exchange refers to the movement and mixing of surface water and groundwater in the benthic surface and shallow subsurface environment of streams (i.e., the hyporheic zone). Hyporheic flow paths, both lateral and vertical, as well as spatial patterns of hyporheic exchange between river, streambed, and aquifer represent the combined effects of hydraulic head differentials produced by surface water and groundwater inflows (i.e., hydraulic forcing) as well as differences in streambed hydraulic conductivity (K) and geomorphic conditions (Krause et al. 2012; Caruso et al. 2016; Schmadel et al. 2017). Vertical expression of groundwater in streams results not only from shallow hyporheic exchange processes, noted previously, but also from regional groundwater inflow (Hyun et al. 2011) that can produce spatially discrete patches of cold water along preferential flow paths and macropore openings (Burkholder et al. 2008; Menchino et al. 2015). These cold-water patches in warm streams can serve as thermal refuges and have become increasingly focused upon when considering projected changes to thermal regimes under rapidly changing climate conditions (Isaak et al. 2012; Kurylyk et al. 2014, 2015a, 2015b); however, much remains to be understood in terms of their physical properties, as well as their spatial and temporal dynamics under changing streamflow, groundwater, and temperature conditions. Research into the controls on hyporheic exchange processes has been done at various scales (e.g., by macro-habitat type, reach, watershed, and basin) and in different topographies (e.g., steep, mountainous terrain vs. low-gradient, unconfined alluvial floodplains). Many of these studies use one and two-dimensional models to help understand hyporheic exchange 34 processes and have consistently identified two major controls to hyporheic exchange: groundwater inflow and geomorphic conditions (Caruso et al. 2016; Schmadel et al. 2017). These studies provide important insights into the physical hydrology of groundwater and surface water interactions in mainstem rivers but do not capture many of the site-level habitat anomalies that occur in riverscapes and are used by fish. By first identifying thermally anomalous fish habitats, likely due to the presence of persistent groundwater exfiltration to the surface, my goal was to collect empirical data that could be used for modelling temperature and hyporheic exchange conditions and controls with specific reference to fish and fish habitat. The objectives of this study were to characterize in situ thermal conditions and groundwater upwelling at discrete thermal refuges, or cold-water patches, as well as to examine changes in response to mainstem stream temperature, air temperature, and stream discharge. Methods Study Location I conducted this study on the Nicola River, in the Southern Interior region of British Columbia (BC), Canada (Figure 2.1). The Nicola River is a sixth order stream, has a watershed area of 7,211 km2, is 188.5 km long (MOE 2022) and has its confluence with the Thompson River at Spences Bridge, which subsequently flows into the Fraser River, approximately 40 km downstream at the town of Lytton. The Nicola River has one mainstem dam which is located at the outlet of Nicola Lake (Figure 2.1). The Nicola Watershed is climatically diverse and includes wet, high-elevation sub-watersheds of the Coast and North Cascades Mountains, to dry, midelevation sub-watersheds to the east, in the rain-shadow of these mountain ranges. Valley bottoms along the mainstem Nicola River exhibit the bunchgrass ecosystems characteristic of 35 semi-arid climates in the region but have slightly more precipitation and are cooler (Lloyd et al. 1990), due to higher elevations. The annual flow regime for the Nicola River has a pattern typical of a snow-dominated watershed with peak flows coinciding with snow melt, between late-April and early-June (Water Survey of Canada 2023). In recent years, fall flooding has increased in frequency and severity (Warkentin 2020) which is more typical of rain-dominated watersheds to the west. The Nicola River exhibits high temperature-sensitivity and has historically been given specific designation to this effect (Reese-Hansen et al. 2012). High summer stream temperatures are common (i.e., >20°C) in the mainstem river and tributaries, with many tributaries exhibiting similar temperature maxima, but higher diel variability than the mainstem Nicola River. This is likely due to the release of epilimnetic flow from the Nicola Lake dam, following thermal stratification in the summer, and this discharge’s higher thermal inertia in relation to its volume. Groundwater-surface water interactions in the Nicola watershed, including hyporheic exchange, groundwater upwelling from aquifers, and small groundwater-fed tributaries, have been found to be important sources of cool-water refuges for salmonids during high summer stream temperatures (Walthers and Nener 1997; Walthers and Nener 2000). 36 Site Selection I installed nested wells at five locations that were identified during the thermal infrared imaging phase of my research (see Chapter 1). Site selection criteria included having prominent cool-water signatures in thermal infrared imaging that appeared to be from groundwater exfiltration (vs. tributaries), being able to establish practical land access to sites for installation and monitoring of equipment, and having sites spaced across the length of the study reach. Site locations are shown in Figure 2.1. Figure 2.1. Location of study area and individual study sites. Inset map shows the location within the Pacific Northwest region of the United States and Canada. 37 Equipment Nested wells consisted of a 3.81 cm (1.50 inch) inside diameter PVC piezometer that was installed in the streambed to a targeted depth of 0.5 m. The bottom of the piezometer was glued and capped and had a 20 cm long slotted-screen segment immediately above, created with a 0.65 mm (0.025 inch) thick standard bandsaw blade. To install the piezometer, I designed a dual-tube installation unit similar in concept to that described by Baxter et al. (2003). The unit consisted of a 5.08 cm (2.00 inch) diameter solid steel drive rod, machined to a point on the bottom and a steel collar welded 5 cm from the top, with the top of the drive rod functioning as the hammer cap. The drive rod fit into a 5.24 cm (2.06 inch - inside diameter) steel casing that was the length of the unmachined segment of the drive rod (1.5 m), between its point and steel collar. The bottom end of the casing was beveled to create a continuously smooth point when the drive rod was inserted in its casing. This allowed the unit to be driven into the streambed with minimal resistance. To install the piezometer, the drive point and casing were hammered into the streambed using a 4.5 kg sledgehammer. Once the targeted depth was achieved (i.e., 0.5 m), the drive point was removed, and the piezometer was inserted into the casing. While placing pressure on the top of the piezometer, the casing was then removed, leaving the piezometer in the streambed at the required depth. The gravel around the piezometer was then tamped by hand to ensure surface water would not infiltrate the piezometer through gaps left by the casing. Following installation, I purged/developed the piezometer by pumping water from it using 13 mm (0.5 inch) standard Waterra tubing with a foot valve (Waterra Pumps Ltd., Mississauga, Ontario, Canada). Piezometers were purged continuously for 10 minutes or until purged water appeared clear. A stilling well was then attached to the piezometer which consisted of a PVC tube of the same diameter that is open at the bottom and positioned slightly above the streambed 38 (Figure 2.2). The stilling well was used to measure stream stage (i.e., the water level of the river at that location, relative to that of the piezometer). With the stilling well in place, I was able to determine the hydraulic head differential in the piezometer. If the water level in the piezometer was higher than that of the stilling well, it would indicate upwelling at the site. If it was below the level of the stilling well, it would indicate downwelling (Baxter and McPhail 1999; Baxter et al. 2003). For comparing upwelling and downwelling between sites and quantifying groundwater-surface water interactions, I calculated Vertical Hydraulic Gradient (VHG) which is a unitless metric that uses the following formula: VHG = Δh / Δl. Where ∆h represents the head differential between the piezometer and the stilling well (i.e., the elevation of the water in the piezometer minus that of the stilling well relative to a local datum), and ∆l represents the depth from the surface of the streambed to the first opening in the perforated segment of the piezometer. Under this formula, upwelling conditions are indicated by positive values and downwelling conditions are indicated by negative values. I monitored water levels in both piezometers and stilling wells from July 11 to October 12, 2017, and July 14 to October 30, 2018. Water levels were monitored at 15-minute intervals throughout the data collection periods using Hobo® U20 series pressure transducer and temperature loggers (Onset, Bourne, Massachusetts, USA). The manufacturer specifies a typical water level Figure 2.2. Streambed groundwater monitoring well diagram (under upwelling conditions), modified from Baxter er al. (2003). 39 accuracy of ± 5 mm for this model. Discrete water level measurements were also collected using a Heron dipper-T electronic water-level meter (Heron Instruments Inc., Dundas, Ontario, Canada). These measurements were based on the height from the top of the PVC piezometer, after installation of equipment, prior to retrieval, and periodically during the data collection period. These measurements were later used to reference water levels recorded by the pressure transducers to a local datum. Figure 2.2 illustrates the typical configuration of my monitoring equipment with variables used to calculate VHG (modified from Baxter et al. 2003). In addition to piezometer and stilling well temperatures, I logged temperature data in 15minute intervals using Onset TidbiT v2 data loggers (Onset, Bourne, Massachusetts, USA) at mainstem river locations adjacent to each site as a reference of background stream temperatures. I checked the calibration of my temperature loggers by submerging them in an ice-water bath to examine the consistency between loggers against the manufacturer’s accuracy specifications of ± 0.2 °C. The manufacturer also notes that loggers will effectively detect changes in temperature at a resolution of 0.02 °C. All loggers were found to be working properly and within an acceptable range when examining their time-series temperature profiles during this calibration check. Some minor variations in temperature profile were noted in the range of 0.3-0.4 °C but it was not expected that the ice-water bath was perfectly uniform. Temperature loggers were deployed in perforated, white PVC housings to protect them from damage and to reduce the effects of radiant heating from the sun (Dunham et al. 2005) and were anchored on top of the stream bed using a PVC coated cable attached to a steel plate. 40 Data Analysis Prior to analysis, I created a time-series data set that included the daily averages of the following environmental covariates: water level (m) and temperature (°C) for both piezometer and stilling well, VHG, mainstem/background stream temperature (°C), air temperature (°C), atmospheric barometric pressure (kPa), stream discharge (m3∙s-1), and a categorical variable called thermal refuge type. I checked the calibration of my pressure transducers, using their average daily water levels and corresponding discrete manual measurements (n=25), by comparing hydraulic head differentials between the two and calculating the average error. Based on this, I calculated an average error of 5.6 mm ± 1.4 mm SE. All statistical analyses were performed in R version 4.1.2 (R Development Core Team 2021). The inferential goals of my study were to examine the effects of environmental covariates, and differentials between time-series measurements, on changes in stream temperature, thermal refuge temperature (i.e., stilling well temperature), groundwater temperature (i.e., piezometer temperature), and VHG using multiple linear regression techniques under maximum likelihood methods. To do this, I plotted covariates against each other and visually checked for correlations of interest. Time-series data were decomposed by applying first-order differencing to remove the effects of trends between consecutive observations (i.e., ΔY = Yt – Yt-1). After fitting models of varied complexity, I compared them using Akaike Information Criterion (AIC) and ranked them according to their AIC score. Model assumptions were also checked to ensure homoscedasticity in variance of my minimally adequate model and were checked for correlations between explanatory variables using Pearson’s correlation coefficient (PCC). I also checked linear regression models for autocorrelation between residuals using a Durbin-Watson (DW) test. 41 Where non-linearity was suspected, I used non-parametric smoothers for covariates of interest using generalized additive models (GAMs) with a Gaussian distribution, using the mgcv package in R (Wood 2011), to examine curvatures in these relationships, if any, using restricted maximum likelihood (REML) methods. GAM models were fitted, and assumptions were checked using the ‘appraise’ function in the r-package, gratia (Simpson 2023) which includes typical diagnostic plots of residuals. Results Site-level summary of temperature and VHG Thermal refuges showed a high degree of variability in VHG and temperature regimes across all sites and between years. Although off-channel to the mainstem Nicola River, thermal refuge temperatures at Site 1 remained close to mainstem temperatures. During 2018, mean thermal refuge temperatures during summer months (i.e., July-September) were higher than mainstem temperatures. I found this conspicuous considering that mean thermal refuge temperatures at this site during the summer of 2017 were slightly cooler. I did, however, observe that preceding my 2018 field season a large-scale flood event during the spring freshet resulted in the formation of a large gravel bar at this site that was not shaded by riparian vegetation. Mean monthly thermal refuge temperatures at this site ranged between 19.1 ± 1.2 SD °C and 10.6 ± 2.0 SD °C between July and October of 2017, and 19.9 ± 2.0 SD °C and 8.1 ± 1.7 SD °C between July and October of 2018 (Appendix A). Piezometer/groundwater temperatures at this site were similar to stilling well temperatures and were generally only slightly cooler in summer months (i.e., July – August) and slightly warmer in fall months (i.e., September – October). At all sites, I noted a general pattern of groundwater temperatures being cooler than thermal refuge 42 temperatures in the summer, with a crossover generally observed in September where groundwater was warmer, relative to thermal refuge temperature. VHGs at this site indicated slight downwelling, predominantly, and ranged between 0.023 and -0.087 with a mean of -0.032 ± 0.034 SD in 2017. For 2018 values were between 0.000 and -0.020 with a mean of -0.009 ± 0.004 SD. Site 2 differed the most compared to the mainstem stream temperatures and displayed strong positive VHG, particularly in 2018. This site was a cold alcove that was off-channel to the mainstem of the Nicola River, had visible seepage and upwelling throughout the study period, and had average temperatures during the summer that were sometimes more than 10°C cooler than mainstem temperatures. This habitat was isolated from the mainstem, apart from its confluence, by a large, extended gravel bar that was deposited during lateral shifting of the river. Mean monthly thermal refuge temperatures ranged between 11.8 ± 1.3 SD °C and 8.2 ± 1.0 SD °C between July and October of 2017, and 9.9 ± 0.2 SD °C and 9.3 ± 0.3 SD °C between July and October of 2018 (Appendix A). VHGs over the same time periods ranged between 0.164 and -0.069 with a mean of 0.096 ± 0.055 SD in 2017, and between 0.290 and 0.220 with a mean of 0.259 ± 0.015 SD in 2018. Groundwater temperatures were slightly cooler than thermal refuge temperatures with a cross-over exhibited in October. VHGs at this site in 2018 appeared to mirror the surface water hydrograph, where changes in stream discharge resulted in an inverse change upwelling. Site 3 was identified as a diffuse hyporheic upwelling within the mainstem of the river during thermal infrared imaging (Chapter 1). Mean monthly thermal refuge temperatures (i.e., stilling well temperatures) remained close to mainstem stream temperatures and ranged between 19.9 ± 1.9 SD °C and 8.1 ± 2.2 SD °C between July and October of 2017, and 21.4 ± 2.3 SD °C 43 and 7.7 ± 1.2 SD °C between July and October of 2018 (Appendix A). Average piezometer temperatures in 2017 remained close to stilling well temperatures but were substantially cooler in 2018 during July and August (Δ = 8.6 °C and 4.9 °C, respectively). VHGs over the same time periods ranged between 0.239 and 0.001 with a mean of 0.092 ± 0.062 SD in 2017, and between 0.031 and -0.014 with a mean of -0.003 ± 0.006 SD in 2018. Upwelling at this site in 2017 displayed a pattern of decreasing VHG from positive values towards zero, following the diminishing limb of the hydrograph towards summer base-flows, post-freshet. Interestingly, this pattern was not observed at this site in 2018, and instead remained near zero. I suspect that this may have been due to fine-scale differences in streambed conditions, as piezometer locations were not exact between years, as well as changes to the streambed during the preceding freshet. Site 4 was a mainstem thermal refuge that appeared to be associated with lateral seepage along the valley wall (i.e., where the river runs along the edge of the valley floor). This site displayed consistent, positive VHGs in both 2017 and 2018 with generally low variability. Mean monthly thermal refuge temperatures at this site ranged between 19.5 ± 2.1 SD °C and 8.8 ± 1.5 SD °C between July and October of 2017, and 20.9 ± 2.5 SD °C and 7.6 ± 1. 3 SD °C between July and October of 2018 (Appendix A). Piezometer temperatures remained cooler than stilling well temperatures during the summer with a cross-over occurring in the fall. VHGs in July through October ranged between 0.017 and 0.037 with a mean of 0.027 ± 0.004 SD in 2017, and between 0.042 and 0.084 with a mean of 0.059 ± 0.009 SD in 2018. Similar to Site 2, but to a lesser degree, more rapid changes in August stream flows were observed to coincide with inverse changes to VHG. Similar to Site 4, Site 5 was a mainstem thermal refuge that appeared to be associated with seepage along the base of the valley wall. This site was abandoned in 2018 after spring 44 freshet resulted in the river shifting laterally in the direction of the opposite bank and partly aggrading the former channel as it shifted. Mean monthly thermal refuge temperatures were similar to mainstem stream temperatures and ranged between 20.9 ± 2.5 SD °C and 7.6 ± 1.3 SD °C between July and October of 2017 (Appendix A). Average piezometer temperatures at this site were substantially cooler than stilling well temperatures during July and August (Δ = 7.0 °C and 4.8 °C, respectively) but were slightly warmer during September and October, 2017. VHGs in July through October ranged between 0.088 and -0.020 with a mean of 0.011 ± 0.023 SD. VHGs at this site stayed near zero or slightly positive through most of the 2017 data collection period but exhibited a marked increase between late August and early September. VHG and temperature patterns and thermal refuges Patterns in VHG by site were quite variable in 2017 but, at some sites, exhibited a pattern of decreasing VHG that coincided with decreasing river discharge, following spring freshet (Figure 2.3). Conversely, in 2018, rapid short-term pulses of river discharge, from dam releases and precipitation events, produced conspicuous changes in VHG that were inversely proportional to changes in discharge (both positive and negative; Figure 2.4). River temperatures and mainstem thermal refuge temperatures (i.e., stilling well temperatures) were most sensitive to changes in air temperature, whereas off-channel thermal refuges and piezometer temperatures were generally less sensitive to changes in air temperature (e.g., Site 2), as is described later (Figures 2.5 and 2.6; Appendix B – Site 2). 45 Figure 2.3. (a) Time-series mean daily stream discharge (Nicola River at Spences Bridge) in relation to (b) mean daily VHG by site in 2017. VHG was highly variable in 2017 but exhibited some general patterns in relation to stream discharge by site. (c) VHG plotted in response to stream discharge by site in 2017. 46 Figure 2.4. (a) Time-series mean daily stream discharge (Nicola River at Spences Bridge) in relation to (b) mean daily VHG by site in 2018. Note the inverse pattern of VHG in response to rapid changes in stream discharge, particularly at Site 2. (c) VHG plotted in response to stream discharge by site in 2018. 47 Mainstem temperature response to environmental covariates Of the explanatory environmental covariates (air temperature, VHG, and stream discharge (Q)), air temperature explained most of the variation in mainstream river temperature response, independently, and across all sites and years (P <0.001; R2 = 0.890) (Appendix C). VHG was only weakly correlated with stream discharge when including all data, however, there were strong correlations at some sites and years, and so I chose to exclude this covariate from the model. For example, at Site 2 in 2018 VHG had a strong negative correlation with mean daily stream discharge (m3∙s-1) (PCC = -0.72), and as will be described later, VHG is not independent and shows a response to changes in stream discharge. Under multiple regression, the simple model of mainstem river temperature response (ΔT.Stream) including first-order differencing of air temperature (ΔT.Air) and stream discharge (ΔQ) as well as a first-order lag of air temperature change (ΔT.Air(t-1)), had the lowest AICc score and explained 84% of the total predictive power of the models included in Table 2.1 (according to its AICc weight score, which is not included in the table). This model is represented by the following equation: ΔT.Stream = -0.09 + (0.25∙ΔT.Air) + (0.16∙ΔT.Air(t-1)) + (-0.09∙ΔQ). Table 2.1. Mainstem stream temperature model summary. "ΔT.Air + ΔT.Air(t-1) + ΔQ" was the most parsimonious model, following removal of VHG as an explanatory variable due to correlation issues between Q and VHG at some sites. Model AICc ΔAICc Adj.R2 p-value ΔT.Air + ΔT.Air(t-1) + ΔQ 196.70 0.713 <0.001 ΔT.Air + ΔT.Air(t-1) 200.03 3.33 0.706 <0.001 ΔT.Air + ΔQ 289.04 92.34 0.540 <0.001 ΔT.Air 294.58 97.88 0.523 <0.001 48 Site-level thermal refuge and groundwater temperature response to environmental covariates Site-level differences in thermal refuge temperature (i.e., stilling well temperature) and groundwater temperature (i.e., piezometer temperature) regimes are conspicuous in ungrouped data and needed to be grouped by site and year to analyze site-level sensitivity of thermal refuges and groundwater to changes in air temperatures (Appendix C). After parsing the data, off-channel sites (Sites 1 and 2) were less sensitive to changes in air temperature in 2017 than mainstem thermal refuges (Sites 3, 4, and 5) that were only slightly less sensitive than the previously noted average mainstem temperatures (Figure 2.5). Coinciding with observed changes in river morphology and formation of exposed gravel bars in 2018, Site 1 exhibited higher thermal sensitivity in relation to air temperature during that year when compared to mainstem stream temperatures (Figure 2.5). Thermal sensitivity in relation to air temperature, however, also increased in 2018 at mainstem thermal refuges (Sites 3 and 4) (Figure 2.5). At Site 2, I observed even less thermal sensitivity in 2018 in relation to air temperature and my simple linear regression showed no significant relationship between these two variables (P = 0.153) (Figure 2.5) as temperatures at this site were extremely stable, with mean monthly temperatures ranging only between 9.3 – 9.9 °C from July through October (Appendix A). Groundwater (i.e., piezometer) temperatures were less sensitive to changes in air temperature (Figure 2.6). Similar to its surface water expression, groundwater at Site 2 showed the least thermal sensitivity in both 2017 and 2018. In contrast, groundwater at Site 4 became very thermally sensitive in 2018 after showing almost no thermal sensitivity in 2017 (Figure 2.6). 49 Figure 2.5. Linear regressions with 95% confidence interval of site-level differences in thermal sensitivity of thermal refuge stilling well temperature (SWTemp) in relation to air temperature (Air Temp) at thermal refuge sites. 2017 data are indicated by solid circles and regression lines, and 2018 data are indicated by open circles and dashed regression lines. All models were highly significant (P = <0.001), except for that of Site 2 in 2018 (P = 0.153) which exhibited no significant relationship between air temperature and thermal refuge temperatures. 50 Figure 2.6. Linear regressions with 95% confidence interval of site-level differences in thermal sensitivity of groundwater / piezometer temperature (PzTemp) in relation to air temperature (Air Temp) at thermal refuge sites in 2017. 2017 data are indicated by solid circles and regression lines, and 2018 data are indicated by open circles and dashed regression lines. All models were highly significant (P = <0.001), except for that of Site 2 in 2017 (P = 0.081) which exhibited no significant relationship between air temperature and thermal refuge groundwater temperatures. 51 As with the mainstem temperature model, I fitted numerous site-level temperature models that included lagged (t-1) air temperature and stream discharge as explanatory variables. For those models that were significant (P <0.05) and showed improved fit with the addition of one or both of the previously noted explanatory variables, details have been included in Table 2.2 and are based on the following equations: ΔTSW/Pz = α + (β1∙ΔTair) + (β2∙ΔTair(t-1)) + (β3∙ΔQ), where ΔTSW/Pz = change in daily mean thermal refuge (stilling well) or groundwater (piezometer) temperatures (°C), respectively; ΔTair = change in mean daily air temperature (°C); ΔTair(t-1) = lagged (t-1) change in mean daily air temperature (°C); and, ΔQ = change in mean daily discharge (m3∙s). Where coefficient values are not included in Table 2.2, that variable was either not significant or did not improve the performance of the site-level model and was not included. Table 2.2. Summary of multi-variate site-level thermal refuge and groundwater temperature flux models by year. β1 represents the air temperature coefficient, β2 represents the lagged (t-1) air temperature coefficient, and β3 represents the mean daily discharge coefficient. ΔAICc represents model improvement from site-level univariate model of air temperature only. Site Year α Thermal Refuge Temperature 1 2017 -0.15 2 2017 0.01 3 2017 -0.07 3 2018 -0.09 4 2017 -0.06 4 2018 -0.08 5 2017 -0.06 Groundwater Temperature 1 2017 -0.15 2 2017 0.01 3 2017 -0.08 3 2018 -0.04 4 2018 -0.08 β1 β2 β3 ΔAICc Adj.R2 p-value 0.08 0.18 0.33 0.29 0.27 0.32 0.28 0.05 0.13 0.15 0.15 0.12 0.13 -0.45 - 5.50 7.39 34.68 44.36 42.97 24.46 19.58 0.24 0.61 0.80 0.75 0.72 0.70 0.61 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 <0.001 0.07 0.015 0.16 0.04 0.10 0.01 0.16 0.04 0.14 -0.52 -0.05 -0.07 -0.05 8.39 9.14 64.85 11.28 69.64 0.25 0.26 0.68 0.20 0.63 <0.001 <0.001 <0.001 <0.001 <0.001 52 Stream discharge and VHG dynamics Using GAMs with a Gaussian distribution under REML methods and the entire 20172018 data sets, including all sites, change in stream discharge was significant in its effects on VHG but had a low coefficient of variation (P = 0.0351, adjusted R2 = 0.00395) due to high variability between years and individual sites. To account for this, I modelled changes in VHG between years, by site for both years, and by site between years. My first model revision included examining the effects produced between years, as 2017 and 2018 had substantial differences in their hydrographs. Changes to stream discharge in 2017 were gradual (Figure 2.3) and did not include the pronounced pulses of discharge that appeared to drive the changes in VHG seen in 2018 (Figure 2.4). By year, the effects of non-site-specific changes in stream discharge on changes in VHG were found to not be significant for the 2017 dataset; however, the 2018 dataset yielded highly significant model terms but still had a low coefficient of variation (P = <0.001, adjusted R2 = 0.0417). This fitted model displayed a simple linear relationship between ΔQ and ΔVHG (effective degrees of freedom (EDF) = 1.003) (Figure 2.7). 53 Figure 2.7. Generalized Additive Model (GAM) of the effects of change in daily stream discharge on change in VHG in 2018, including all sites. Plot includes 95% confident intervals, including standard error of model terms and intercept. Note that although GAMs can be used for modelling non-linear relationships between model variables, this model has an effective degrees of freedom value of 1.003, indicating a linear relationship between ΔQ and ΔVHG. Site-level modelling revealed that, among the five sites, only Sites 2 and 3 showed significant correlations between variables, but both models still explained little of the variation in VHG response (Site 2, P = <0.001, adjusted R2 = 0.0681; Site 3, P = 0.021, adjusted R2 = 0.0521). I then refined my models for Sites 2 and 3 to examine site-level differences between years. For Site 2 in 2017, model terms were not found to be significant; however, at Site 2 in 2018 change in discharge was highly significant in its effects on VHG and the model’s coefficient of variation was much better (P = <0.001, adjusted R2 = 0.528). Site 2 was, by far, the site of strongest upwelling and the VHG data from 2018 appeared to be highly sensitive to changes in stream discharge (Figure 2.4). The plotted model displays substantial 54 curvature/wiggliness in the relationship between ΔQ and ΔVHG (EDF = 4.292) towards more extreme daily changes in discharge but displays the linear relationship (Figure 2.7) between approximately ±1 m3/s ΔQ. The model intercept was not significant (P = 0.878), which highlights the increasing effects of changes in stream discharge on ΔVHG as one moved away from zero ΔQ. A plot of this model has been included as Figure 2.8 with 95% confident intervals, including standard error of model terms and intercept. Figure 2.8. Generalized Additive Model of the effects of change in daily stream discharge on change in VHG in 2018 at Site 2 only. Plot includes 95% confident intervals, including standard error of model terms and intercept. Note that this model, fitted to data for Site 2 only, retains the linear relationship between approximately ±1 m3/s ΔQ, but decreases in confidence towards more extreme changes in daily discharge. Changes in discharge at this site were most strongly correlated with inverse changes in groundwater upwelling, according to its VHG. 55 For Site 3, parsing this dataset into separate years improved significance in model term ΔQ and also its coefficient of variation (2017, P = <0.001, adjusted R2 = 0.151; 2018, P = <0.001, adjusted R2 = 0.352). Unlike Site 2, Site 3 showed positive effects to VHG in response to increases in stream discharge. Plots of both models have been included as Figure 2.9 with 95% confident intervals, including standard error of model terms and intercept. Figure 2.9. Generalized Additive Model of the effects of change in daily stream discharge on change in VHG at Site 3 for years 2017 (a) and 2018 (b). Plot includes 95% confident intervals, including standard error of model terms and intercept. Note that, unlike the Site 2 model included in Figure 2.8, increases in stream discharge are generally associated with increases in VHG. Discussion Temperature regimes between thermal refuges varied, and all thermal refuges displayed degrees of anomaly when compared to mainstem river temperatures. The most dramatic differences in thermal refuge temperature relative to mainstem temperatures were noted in the 56 piezometers and in thermal refuges that were isolated from, or off-channel to, mainstem habitats (e.g., Site 2). Temperatures at mainstem thermal refuges were generally close to ambient stream temperatures, likely due to rapid mixing of exfiltrated hyporheic water with surface water. This suggests that temperature differences experienced or selected for by fish in mainstem thermal refuges would be limited to the near-benthic environments only during times of positive VHG. This is not to imply that these thermal refuges are of less importance, however, as their proximity to non-thermal refuges create heterogeneity in the stream thermal environment and may allow fish to more easily exploit mainstem habitats (Brewitt and Danner 2014). Mainstem habitats have been shown to still be important feeding areas for juvenile salmonids during periods of high stream temperatures (Brewitt et al. 2017), and the combination of mainstem habitats and thermal refuges in proximity to each other allow fish to manage trade-offs between thermal and trophic resources therein optimizing feeding, digestion, and assimilation (Armstrong et al. 2013; Baldock 2016). Mainstem stream temperatures at non-thermal refuges were most strongly affected by changes in air temperature and showed a lagged effect to preceding atmospheric conditions. The high degree of mainstem temperature sensitivity to air temperature was concerning, especially during more extreme atmospheric events, but fits within the current understanding of temperature fluxes on alluvial streams. Caissie (2006) describes heat exchange processes at the air/water interface, such as absorption of solar radiation and convective heat exchange from wind, as dominating the overall heat fluxes in larger channels when compared to streambed heat fluxes. Diel temperature variations are also noted to be greatest in mid-sized rivers that are wide (i.e., > 50 m) and shallow (i.e., generally < 1.5 m) (Caisse 2006), like the Nicola River. The high degree of mainstem temperature sensitivity is especially concerning considering stream temperature 57 responses to forest and riparian clearing (Rayne et al. 2008; Cunningham et al. 2023), which is widespread in the Nicola River watershed (Lewis 2016) along with increased summer air temperatures (Warkentin et al. 2022). In contrast to air temperature, the heating effects of solar radiation on streams may be effectively managed by protecting and promoting riparian vegetation. Alteration of shade has significant effects on summer stream temperatures (Woltemade and Hawkins 2016; Roon et al. 2021) and, by providing more of a suitable type (Dugdale et al. 2018), can substantially cool streams, and thermal refuges specifically (Ebersole et al. 2003a). The effects of shade on stream temperatures are reduced, however, in larger channels (Caisse 2006; Quinn and Wright-Stow 2008) as a function of the ratio of tree canopy height relative to channel width (Rutherford et al. 2018). This may limit the effectiveness of restoring riparian cover to reduce summer stream temperatures in middle to lower reaches of the Nicola River; however, the shape of riparian canopy (e.g., one that overhangs the stream) also influences the interception of solar radiation and may have important effects in shading nearstreambank environments (Rutherford et al. 2018) which is where I found thermal refuges often to be located. The multiple regression analysis of mainstem temperature response to environmental covariates showed a small, but significant, decrease in mainstem temperatures in response to increases in stream discharge. These findings are in keeping with other literature that report decreases in stream temperature with increased stream discharge (Woltemade and Hawkins 2016); however, it is important to note that the source of water may be a key factor in these scenarios when comparing discharge from snowmelt versus controlled releases from dams and reservoirs, or a combination of both, as is the case in the Nicola Watershed. Poole and Berman (2001) note that the thermal inertia of higher flows can allow a stream to resist changes in 58 temperature; however, I would add to this by noting that this can act positively or negatively on thermal regimes depending on background stream temperatures and the sources of advective heat fluxes. The rapid changes in discharge resulting in inverse changes to VHG at thermal refuges has important implications in the management of regulated rivers, and for rivers with shifting hydrological regimes due to climate and landscape change-induced stochasticity. Other studies note general patterns of upwelling in the hyporheic zone during summer, transitioning to downwelling in the fall, winter, and spring in alluvial stream channels similar to the Nicola River (Harris and Peterson 2020). These patterns may generally be true for mainstem hyporheic exchange processes, but this was not observed with any consistency at thermal refuges. Only two of five sites in 2017 displayed diminishing, positive VHGs from summer to fall, and this was not observed at all in 2018. Instead, VHGs tended to fluctuate around a static baseline that was either neutral or positive, especially in 2018; however, I only selected sites that exhibited patterns indicative of groundwater upwelling from my thermal infrared imaging, so conclusions should be limited to these thermal anomalies alone. The variability observed at different sites and between years should also impart some caution in generalizing about hyporheic exchange processes at thermal refuges. In general, stream discharge maxima and minima, and their corresponding river stages, were not strongly correlated with greater differentials in VHG. This observation is consistent with other studies that have found that hyporheic exchange is not strongly affected by stream discharge (Wondzell 2006; Schmadel et al. 2017; Harris and Peterson 2020) despite the pressure differentials created by differences in head elevations between stream and aquifer, but that it is driven more by morphological controls and groundwater input (Valett et al. 1994; Wondzell 59 2006; Schmadel et al. 2017). Harris and Peterson (2020) note that higher stream discharges also produce higher current velocities, and that increased current velocity results in a fluid-lift force that can counteract the effects of the downward vertical compression created through elevated head forces. VHG changed significantly in response to more rapid changes in stream discharge in 2018, and this correlation was particularly strong at the most prominent thermal refuge, Site 2. This observation is similar to that of McCallum et al. (2013) in their description of flow pulses resulting in short-duration shifts of streams from gaining to losing (i.e., positive to negative VHG) in response to high river stage events. I expect that the inverse response in VHG to rapid changes in stream discharge is due in part to the lagged response of groundwater elevations in relation to stream stage under different hydraulic conductivities, as both stream stage (stilling well water level) and groundwater elevation (piezometer water level) were found to respond to changes in stream discharge (Appendix D). This site being off-channel, it is also plausible that VHG responded without the counter-acting effects of high current velocity fluid-lift forces, as described previously for mainstem habitats. The inconsistencies observed between sites reflects the real-world nature of streams and streambeds being heterogeneous in morphology and sediments. This heterogeneity has been shown to affect hyporheic exchange processes and the propagation of pressure as created by differences in hydraulic head between stream and aquifer (Yeh et al. 2009). Simulation of these processes typically assume homogenous sediments (Welch et al. 2013; McCallum and Shanafield 2016), so my work provides empirical examples of variability in hyporheic exchange responses to changes in stream discharge under heterogeneous streambed sediments and morphological features found at thermal refuges. 60 The pulses of stream discharge in 2018 (Figure 2.4) were due to both atmospheric events and planned dam releases. The pulses seen between mid-August and early September originated from releases at the Nicola Lake dam which were intended to boost instream flows for inmigrating Chinook Salmon, at the request of Fisheries and Oceans Canada (J. Ball, personal communication, February 1, 2023). The pulses seen in late September through October of 2018 did not originate from dam releases but were from atmospheric events (typical for that time of year) as evidence by hydrometric data for the unregulated Coldwater River - Station 08FG048, 2018 (Water Survey of Canada 2023), a major tributary to the Nicola River. Although dam releases in late-August are intended to support salmon in-migration, the timing of this results in potentially adverse effects to groundwater conditions at thermal refuges, through reduced VHG, during a time of year when mean stream temperatures are close to 20°C (Table 2.1). Research suggests that under small, short-term stage fluctuations, water infiltrating the streambed can be exfiltrated back to the surface quite quickly, following recession of flows; however, under stream stage fluctuations from larger discharge events (e.g., spring freshet), water entering the aquifer may persist there for longer, resulting in a less immediate exfiltration response under receding flows (McCallum and Shanafield 2016). The combination of advective heat fluxes from releasing pulses of warm, epilimnetic water from Nicola Lake combined with potential for reduced groundwater upwelling at downstream thermal refuges warrants thoughtful decisions regarding environmental flows during periods of high ambient stream temperatures. The relative influence of stream discharge on a stream’s overall temperature regime increases in larger, fifth order or greater streams, like the Nicola River (Poole and Berman 2001). In addition to temperature, in snow-dominated watersheds of semi-arid interior regions of the Pacific Northwest, with hot summers and cool 61 winters, late summer months tend to be a time with predictable flows that allow for recovery of aquatic biota between spring freshet and increasingly stochastic fall and winter flood events (Yarnell et al. 2010). Increasing stream discharge pulse frequency during this time through anthropogenic means (as was seen in 2018), although well-intentioned, has the potential to disrupt aquatic ecosystems during critical, and typically stable, rearing periods for juvenile salmonids. Additionally, the effectiveness of releasing flow pulses from dams to improve upstream migration of Chinook Salmon remains unsupported by research (Hasler et al. 2014; Peterson et al. 2017). This study demonstrates that thermal refuges in alluvial stream channels of snowdominated, semi-arid watersheds are highly variable in their thermal regimes but do provide anomalous conditions when compared to mainstem habitats that can be favorable to cool-water fish species. These habitats can be highly discrete, thermally, and hydrologically, and vary significantly in their sensitivity to changes in air temperature and stream discharge. To conserve the conditions that make these habitats suitable for cool-water fish species, such as Pacific salmon and steelhead, efforts should be made to reduce the primary heat fluxes that cause mainstem and thermal refuge warming, namely exposure to solar radiation through stream shading as well as unnatural manipulations to stream discharge. As climate change and habitat alterations apply increasing stress to aquatic ecosystems, preserving and augmenting thermal refuges through restoration activities and effective management (Kurylyk et al. 2015a) will become increasingly important. Water managers of regulated rivers will also need to consider the timing and rates of discharge changes more carefully from dams in light of natural flow regimes and should be cautious of impacts to hyporheic exchange processes at thermal refuges during peak temperatures and diminishing flows in the summer. 62 CHAPTER 3 Diel horizontal migration of stream-dwelling juvenile Pacific salmonids (Oncorhynchus spp.) between mainstem and thermal refuge habitats 63 Abstract Stream-rearing salmon and steelhead have been shown to selectively exploit heterogeneity in thermal and physical environments through diel horizontal migration (DHM). These movements become a critical part of their life history and aid in feeding, metabolism, rest, and predator avoidance. In salmon streams of semi-arid environments these migrations also allow fish to access thermal refuges when mainstem stream temperatures exceed critical thresholds for growth and survival. I used passive integrated transponder tags and antenna arrays at selected thermal refuges to track DHM of juvenile Chinook Salmon (Oncorhynchus tshawytscha), Coho Salmon (O. kisutch), and Steelhead (O. mykiss) in response to mainstem temperature, thermal refuge temperature, and photoperiod in the Nicola River, British Columbia, Canada. I used a Poisson Generalized Additive Model to predict the effects of environmental covariates on DHM. In addition to this, I fitted a thermal refuge occupancy model for fish that displayed multi-day patterns of DHM using a logistic Generalized Additive Mixed Model. I found that, although fish differentially exploited the cooler temperatures found in thermal refuges during periods of high mainstem stream temperatures, stream temperature was generally a poor predictor of DHM and thermal refuge occupancy. The time of day (in relation to photoperiod) was the strongest predictor of DHM and thermal refuge occupancy. Fish were generally found to migrate at sunrise and sunset and remained in thermal refuges during daylight hours. These migrations continued into the fall, even as mainstem stream temperatures dropped below critical thresholds. This study brings new insights into the environmental cues that trigger DHM of juvenile salmon in streams and their use of thermal and physical habitat heterogeneity. 64 Introduction Pacific salmon (Oncorhynchus spp.) and steelhead (Oncorhynchus mykiss) are cold-water, ectothermic fishes that are highly sensitive to changes in temperature at the individual level (McCullough 1999; Quigley and Hinch 2005; Farrell et al. 2008), but also in terms of population-level impacts from changes in stream temperature regimes (Mantua et al. 2010; Mauger et al. 2017). High water temperatures have been shown to adversely affect growth and survival of juvenile salmon, both in laboratory experiments (Brett 1952; Perry and Plumb 2015) and in the field (Nielsen and Lisle 1994; Davidson et al. 2010), but effects of temperature on in situ growth and survival are often confounded by interactions with environmental covariates (Bal et al. 2011; Campbell et al. 2020; Lusardi et al. 2020). Advancements in temperature sensing technology, including thermal infrared (TIR) remote sensing, have revealed remarkable heterogeneity in stream thermal environments (Dugdale et al. 2019; Chapter 1), which have been shown to be differentially exploited by rearing juvenile salmon and steelhead through diel horizontal migration (DHM) to optimize feeding and metabolism (Armstrong et al. 2013; Brewitt et al. 2017), and to ensure survival (Breau et al. 2007; Breau et al. 2011). The effects of temperature in producing individual and population-level responses in salmon and steelhead (and other ectotherms, for that matter) dominate the literature when compared to other environmental variables (Wade et al. 2013; Carlson et al. 2017). This is not surprising in light of anthropogenic changes to global temperature regimes and degradation of thermal environments in many streams; however, aquatic environments are influenced by numerous external factors, beyond temperature, that trigger movement in fish (Scheuerell and Schindler 2003). Nathan et al. (2008) provide a unified approach to studying the movement ecology of organisms where movements are reduced to a series of steps and stops, driven by 65 external factors acting upon the individual, their internal state, as well as their motion and navigation capacities. This approach reveals some major challenges in studying the movement ecology of juvenile salmon, as external and internal factors that influence movement need to be characterized at fine-scales, as fine-scale heterogeneity has been shown to be a key driver in behavioural thermoregulation in ectotherms (Sears and Angilletta 2015). Thermal refuges in streams have become a topic of great interest in recent decades and include numerous similar definitions (Ebersole et al. 2003a; Torgersen et al. 2012; Dugdale et al. 2013 and 2015; Fullerton et al. 2018) that can generally be summarized as cool-water habitats that are used by fish during periods of high ambient stream temperatures. There is strong empirical evidence that fish occupy thermal refugia during periods of high mainstem stream temperatures (Ebersole et al. 2003b; Breau et al. 2007; Breau et al. 2011; Brewitt and Danner 2014) and that fish will move to mainstem habitats at night to feed (Armstrong et al. 2013). Adding complexity to this understanding is the recognition that the thermal regimes of rivers vary based on numerous factors, both natural and anthropogenic (Cassie 2006). Stream thermal regimes vary naturally based on latitude and prevailing climate, which alters how one defines and understands thermal refuges. For example, at northern latitudes, juvenile Coho Salmon (O. kisutch) have been shown to migrate to warm water habitats, from cold mainstem habitats, to improve metabolism and food assimilation following mainstem feeding forays (Armstrong et al. 2013; Armstrong and Schindler 2013). At southern latitudes, the timing of movements remain similar, however, they do so in response to, essentially, the opposite pattern in environmental variables, where juvenile salmonids feed from warm mainstem food sources while exploiting cool-water refuges to meet metabolic needs (Brewitt and Danner 2014; Brewitt et al. 2017). 66 The study of habitat use by juvenile salmonids in response to spatiotemporal differences in thermal environments is growing and remains a relevant topic in the frontiers of salmon ecology research. Current literature has revealed that fish behaviour in response to individual and environmental covariates in thermally heterogeneous environments are remarkably complex. My research goals were to characterize patterns of diel horizontal movements in juvenile Chinook Salmon (O. tshawytscha), Coho Salmon, and steelhead (hereafter referred to as my focal species) in thermal refuges of a temperature-sensitive, semi-arid, alluvial stream environment. Of the populations represented by my focal species, Interior Fraser Coho Salmon has been assessed as Threatened (2016) by the Committee on the Status of Endangered Wildlife in Canada (COSEWIC), and both Thompson Steelhead and the Lower Thompson stream-type spring Chinook have been assessed as Endangered (Government of Canada 2022). None of these three populations are currently protected under Canada’s Species at Risk Act. Methods Study Location I conducted this study on the Nicola River, in the Southern Interior region of British Columbia, Canada (Figure 3.1). The Nicola River has a watershed area of 7,211 km2, is 188.5 km long (MOE 2022) and has its confluence with the Thompson River at Spences Bridge, which subsequently flows into the Fraser River, approximately 40 km downstream at the town of Lytton. The Nicola River Watershed is climatically diverse and includes wet, high-elevation subwatersheds of the Coast and North Cascades Mountains, to dry, mid-elevation sub-watersheds to the east, in the rain-shadow of these mountain ranges. Valley bottoms along the mainstem Nicola River exhibit the bunchgrass ecosystems characteristic of semi-arid climates in the region but have slightly more precipitation and are cooler (Lloyd et al. 1990), due to higher elevations. The 67 annual flow regime for the Nicola River has a pattern typical of a snow-dominated watershed with peak flows coinciding with snow melt, between late-April and early-June (Water Survey of Canada 2023). In recent years, fall flooding has increased in frequency and severity (Warkentin 2020) which is more typical in rain-dominated watersheds to the west. The Nicola River has historically been classified as a temperature-sensitive stream (Reese-Hansen et al. 2012) under British Columbia’s Forest Practices Code and regularly experiences mainstem river temperatures higher than 25 °C in summer months (Willms 2024, unpublished data; Kosakoski and Hamilton 1982; Walthers and Nener 1997). Anadromous salmon migration is generally unimpeded by dams and other anthropogenic features in the Nicola River and its major salmon-producing tributaries. However, there is one mainstem dam on the Nicola River, at Nicola Lake, that releases epilimnetic flow during the summer, following thermal stratification. Increased demand for water during the summer months, coupled with major changes to the hydrology of the watershed (e.g., large-scale salvage logging and fires) have resulted in water scarcity issues, conflicts amongst water users and regulatory agencies, and potentially adverse conditions for native fish populations. 68 Figure 3.1. Location of study area and individual study sites in the Interior of British Columbia, Canada. Inset map shows the location within the Pacific Northwest region of Canada and the United States. Site Selection To identify potential thermal refuges, I used Remotely Piloted Aircraft System (RPAS)based thermal infrared imaging techniques. The imager was a ZenMuse XT, 640 x 512 resolution, 13 mm lens with a 30 Hz refresh rate (DJI, Shenzhen, Nanshan District, China), and was attached to a Matrice 200 RPAS (DJI, Shenzhen, Nanshan District, China). The imager is built on a gimbal which reduces vibration and maintains image geometry during flight. I completed the initial imaging of my study reach in the summers of 2016 and 2017 with a lower resolution 69 camera (Chapter 1), but re-imaged selected sites in 2020 with the higher resolution system following major flood events in 2017 and 2018, and concurrent with collection of passive integrated transponder (PIT) tag data described in this chapter. Image locations that displayed cool-water signatures and were associated with off-channel habitats (e.g., alcoves, side-channels and tributaries) were further investigated in the field, using snorkel surveys and visual observations from the stream bank, for presence of focal fish species during high mainstem stream temperatures (≥ 20°C). Sites were also assessed as to their suitability for installation of PIT antenna arrays. Having four PIT antennae readers available for the study, I chose and set up three sites in 2020 and four sites in 2021. Fish Capture Fish were captured using two-person crew backpack electrofishing techniques with a Smith-Root LR-24 backpack electrofisher (Smith-Root Inc., Vancouver, Washington, USA) and a dip-net. Electrofishing was conducted in 1-2 passes in selected thermal refuges between 10:00 and 15:00 when it was anticipated that focal species would be occupying these sites. Fish collection was conducted between 1-3 times at each site over the months of July and August during the summers of 2020 and 2021. A maximum of 50 individuals of focal species were captured and tagged per site per year; however, I also ceased fish collection at sites once I had numerous tagged fish exhibiting diel migrations to avoid potential disruption of these movements and loss of data. Enumeration, Tagging and Tracking To track fish movement patterns associated with thermal refuges and to identify recaptured fish during successive sampling events, I used 12.0 mm x 2.12 mm half-duplex (HDX) PIT tags (Oregon RFID, Portland, Oregon, USA). PIT tagging is non-lethal technique 70 that does not normally result in adverse effects to the fish (Tiffan et al. 2015; Acolas 2007). After capture, fish were anesthetized in a solution of tricaine methanesulfonate (MS-222; 60 mg · L–1) buffered with sodium bicarbonate (Neifer and Stamper 2009; Conrad et al. 2016). Fish were identified, scanned for recaptures, and measured to the nearest 1 mm fork-length (FL). PIT tags were then inserted manually into the peritoneal cavity of captured individuals, of my focal species, that were ≥ 55 mm FL (Achord et al. 1996; Acolas 2007), excluding any that exhibited visually poor body condition. Insertions were made by making a small (3-4 mm), un-sutured incision with a scalpel on the ventral side of the body, anterior to the pelvic fins (Gries and Letcher 2002; Conrad et al. 2016), and by inserting and seating the tag by hand. Manually implanting PIT tags has been shown as a viable alternative to PIT tag injectors that promotes high survival and low tag-loss/shedding (Gries and Letcher 2002). After tagging, fish were moved to an aerated bucket where they were left to recover prior to release. I used a combination of one single and three multi-HDX antenna readers (Oregon RFID, Portland, Oregon, USA) across four sites that were powered by 12 V batteries. During the summer of 2020, batteries were exchanged manually on a weekly basis to avoid power loss and interruptions to data collection; in 2021 I installed 160-watt solar panels at all sites to charge batteries in situ and maintain continuous data collection. PIT antenna arrays were constructed using a loop of wire that was installed along the wetted perimeter of the channel in cross-section, according to criteria described by the manufacturer. Antennae at multi-reader sites were installed in series (2 or 3) at thermal refuges near its confluence with mainstem habitats. Having antennae in series allowed me to determine direction of travel during diel migrations. Detections logged on the downstream antennae followed by upstream antenna indicate immigration from mainstem habitats to thermal refuges, and the opposite sequence indicating emigration out of thermal 71 refuges. I used a single reader at one location that was a small (< 10 m2) plume of cool water spilling out of a beaver pond. The site was too small for multiple antennae so, instead, a single antennae was installed to detect tagged fish while they occupied this habitat. Antenna arrays and readers were tested and downloaded weekly to ensure they were tuned and working effectively. After removing my PIT antennae and readers following the first field season (2020), I surveyed all sites with a Biomark mobile antenna and reader (Biomark, Boise, Idaho, USA) for the presence of “ghost tags” that may have been lost due to fish mortality or tag shedding. Temperature Monitoring I monitored temperature at 15 minute intervals in thermal refuges and adjacent mainstem habitats throughout the study period using Onset TidbiT v2 data loggers (Onset, Bourne, Massachusetts, USA). Data loggers were tested together for calibration in an ice-water bath, prior to being deployed in the field, and were accurate to ± 0.2 °C according to manufacturer specifications. To protect the data loggers and reduce the effects of radiant heating from the sun, temperature loggers were housed in a protective, perforated, PVC container as described by Dunham et al. (2005) and anchored in the stream on top of the stream bed. Data Analysis Fish collection data, PIT detections, mainstem temperature and thermal refuge temperature data were imported to and compiled in a Microsoft Access database (Microsoft, Redmond, Washington, USA). Data queries were used to match all PIT detections with timeseries of mainstem and thermal refuge temperature covariates. From this I was able to produce data frames with individual detection records of PIT tag identifiers, site, species, FL, date, time and associated mainstem and thermal refuge temperatures. Diel migration movements were added to my database as a subset of detections that included immigrations and emigrations. 72 All statistical analyses were performed in R version 4.1.2 (R Development Core Team 2021). I modelled the modal distribution of diel detection times by site and species using the multimode package in R (Ameijeiras-Alonso 2021) to test for the presence of multiple modes, their location (i.e., peak detection times), and the location of anti-modes (i.e., least detection times). Significance was tested using the Ameijeiras-Alonso et al. (2019) excess mass test. Detection times were converted to continuous numerical values between 0 and 24 and plotted against mainstem temperature and thermal refuge temperature to visualize patterns and interactions amongst covariates. I also plotted detection times across 2020 and 2021 by sites with daily sunrise and sunset events to visualize the effects of photoperiod on fish movement. To characterize DHM patterns and occupancy in thermal refuges, I fitted two models, recognizing that organismal movement consists of a series of steps and stops driven by an organism’s internal state in combination with external factors (Nathan et al. 2008). To model timing of DHM in relation to mainstem and thermal refuge temperature covariates, I used my bulk detection time data which includes timing of all movement events regardless of directionality or thermal refuge site occupancy. I binned and counted the frequency of detection times in hourly intervals and grouped temperature covariates by degree Celsius integers. To this data, I fit generalized additive models (GAMs) with a Poisson distribution and a log link function using the mgcv package in R (Wood 2011), as the relationship between detection frequency and time was clearly non-linear. Within my GAMs, I compared two different model structures: one with individual smoother terms for model covariates and one with a single common global smoother for interaction among model covariates. I compared models using Akaike Information Criterion (AIC) and, as it is not recommended that AIC scores alone be used for GAM selection (Pedersen 2019), I also considered the inferential goals of my study, my knowledge of system, 73 and the predictive ability of each model structure. To test for the presence of autocorrelation in model residuals, I used a Durbin-Watson test, where values of 2 ± 0.5 indicate no issues with autocorrelation. For my second model, I used my migration data to construct a binary dataset of thermal refuge site occupancy for individual fish that displayed multi-day patterns of DHM between mainstem habitats and thermal refuges. Thermal refuge occupancy, as my response variable, was aligned with mainstem and thermal refuge time-series temperature covariates (i.e., 15 min intervals) and were assigned 1’s following immigration (i.e., occupancy in thermal refuge = “true”) and 0’s following emigration (i.e., occupancy in thermal refuge = “false”). I fitted a generalized additive mixed model (GAMM) with a Bernoulli distribution and logit link function to this data using the gamm4 package in R (Wood 2020) after adding an autocorrelation covariate (lag = 1) from my occupancy term. As with my previously described Poisson GAM, I compared different model structures within my logistic GAMM, again comparing individual smoother terms for each covariate to a single, common global smoother of interaction effects. The fit of separate model structures were compared using the area under the curve (AUC) test statistic after performing receiver operating characteristics (ROC) analysis in the pROC package (Robin et al. 2011) in R. The predictive ability of each model was tested by randomly partitioning my data into a training data set (70%), for training the model, and a test data set (30%), which is excluded from model training and is used for my predictive check. Models with values closest to 1 indicated the best predictive ability. 74 Results Thermal Refuge Selection and Temperature Summary From my thermal infrared surveys, I selected four sites of different thermal refuge habitat types (Figure 3.2). Thermal infrared images of each site and differences in stream temperature at thermal refuges versus mainstem habitats are shown in Figures 3.2 and 3.3, respectively. Site 1, at the confluence of Clapperton Creek with the Nicola River, was characteristic of the cool-water plumes created by tributary streams in mainstem habitats. Mean diel stream temperature maxima for the month of August at Site 1 were 2.3°C (SD = 1.1°C) cooler in 2020 and 1.6 °C (SD = 1.1°C) cooler in 2021 than adjacent mainstem habitat. Site 2 consisted of cool water flowing out of a beaver dam complex adjacent to the mainstem of the Nicola River. The source of water is likely groundwater-influenced but may originate from irrigation supply, as I observed almost complete cessation of flows following irrigation shut-off in the fall. During the summer of 2020, water flowing out of the beaver pond was entering a side-channel adjacent to the Nicola River, which allowed for set-up of paired antennae. However, shortly after re-installing antennae in the same location in 2021 the beaver plugged this outlet, leaving the 2020 side-channel dry and causing the water spill over an earthen segment of the beaver pond, creating a much smaller thermal refuge (approximately 10 m2) downstream, along the margin of the main channel. There was not enough room in this location for paired antennae so I shifted my approach to installing a single antenna in the outfall of cool water that would detect fish occupying the site, rather than migrations between. I refer to this site as “Site 2S”, the “S” indicating the single antenna reader. Mean diel stream temperature maxima for the month of August at Site 2/Site 2S were 1.9°C (SD = 0.4°C) cooler in 2020 and 0.4 °C (SD = 0.8°C) cooler in 2021 than adjacent mainstem habitat. 75 Site 3 is a lateral seep that exhibited strong positive groundwater gradients throughout the study and was clearly not influenced by fluctuations in surface water temperatures. Mean diel stream temperature maxima for the month of August at Site 3 were 11.5°C (SD = 1.9°C) cooler in 2020 and 9.3 °C (SD = 2.6°C) cooler in 2021 than adjacent mainstem habitat. Site 4 is a cold alcove-type thermal refuge that was created following overbank flooding and scour in the spring of 2017. The habitat consists of an approximately 600 m2 off-channel pond and channel that exfiltrates groundwater to the mainstem Nicola River, downstream of its confluence with Guichon Creek (Figure 3.1). This site was added in 2021 after visual observation of abundant salmon fry in the pond. Mean diel stream temperature maxima for the month of August at Site 4 was 0.5°C (SD = 1.3°C) cooler in 2021 than adjacent mainstem habitat. 76 Figure 3.2. Thermal infrared images showing heterogeneity of surface water temperatures between thermal refuge sites and the mainstem Nicola River: (A) Site 1 - at the confluence of Clapperton Creek and the Nicola River; (B) Site 2 – a cool-water outfall from a beaver pond adjacent to the Nicola River; (C) Site 3 – lateral seep of cool ground water and associate offchannel habitat to the Nicola River; and, (D) Site 4 – groundwater-fed alcove created by overbank flow during 2017 freshet on the Nicola River, downstream of Guichon Creek. 77 Figure 3.3. Diel temperature maxima (°C) across all sites during 2020 and 2021 data collection periods. Blue data series indicate thermal refuges while red data series (open points) indicate adjacent Nicola River mainstem. 78 Fish Collection, Tagging, Detection, and Migration Summary Between the months of July and August for 2020 and 2021, I captured 289 fish of my focal species across four thermal refuge sites, including: 55 Chinook Salmon (CH), 192 Coho Salmon (CO), and 42 Rainbow Trout (RB). Of these, I PIT-tagged 145 fish, including: 43 CH, 64 CO, and 38 RB. Mean FLs of captured fish were significantly different between all three species (Student’s t-test; P < 0.001) with RB being largest and most variable (mean = 119.67 mm; SD = 47.73 mm), CO being smallest (mean = 57.31 mm; SD = 8.75 mm), and CH being slightly larger than CO (mean = 68.58 mm; SD = 10.12) (Figure 3.4). The higher degree of variability in size in RB vs. CH and CO is likely due to the longer freshwater residency period of anadromous Steelhead (i.e., more age-classes) in combination with non-anadromous RB also being represented in the sample. Figure 3.4. Boxplots of fork lengths of focal species captured during project sampling. 79 In total, I logged 37,908 PIT tag detections for 2020 and 2021 with proportions of detections varying slightly by species and site. Of the individuals that were PIT-tagged in thermal refuges, 48.3% (n = 70) were later detected at PIT-antennae. By species, 51.2% of CH (n = 22), 45.3% of CO (n = 29), and 50.0% of RB (n = 19) were detected again after tagging. Subsequent detection of tagged fish by site included: 47.9% at Site 1 (n = 23); 64% at Site 2/Site 2S (n = 16); 54.5% at Site 3 (n = 12); and 38.0% at Site 4 (n=19). In addition to this, two fish tagged at Site 1 (a 71 mm F.L. CH and a 98 mm F.L. RB) in 2021 were later detected downstream at Site 2S. During my check for “ghost tags” following the 2020 field season, none were found, indicating that tags were likely not being shed after implantation and that mortalities due to tagging were likely negligible. Diel detection frequency was significantly bimodal for all multi-antenna sites (P < .001). Table 3.1 provides mode times, indicating peak movement between thermal refuges and mainstem habitats, as well as each corresponding anti-mode, indicating time of least movement between thermal refuges and mainstem habitats. Pooled detection time frequencies between all antennae at each site have been included in Figure 3.5. The frequency of detections across all multi-antenna sites reveals similar patterns of movement by my focal species that correspond with crepuscular activity. This pattern is also consistent with the detection frequencies at Site 2S, where detections were recorded while fish occupied this site as indicated by the consistency of detections during daylight hours and the general absence of detections at night. Table 3.1. Peak (modes) and least (anti-mode) detection times (24 hr.) by site. Site AM Mode Anti-Mode (Daytime) PM Mode 1 04:17 10:04 20:59 2 04:03 13:43 20:41 3 05:36 14:11 20:19 4 04:23 10:15 21:19 80 Figure 3.5. Diel detection frequencies of fish at all sites across 2020 and 2021. Site 2S is represented in solid grey within the same panel as Site 2. Note that Site 2S is a single antenna reader that detected occupancy at that thermal refuge. From my bulk detection data at all multi-antenna sites, I was able to identify 494 migration events (i.e., immigration to thermal refuge or emigration from thermal refuge to the river mainstem). At Site 1, I was only able to identify 180 migration events, 97 emigrations and 83 immigrations, from 2397 total detections. Of these migrations there was no discernable pattern in immigration vs. emigration at modal detection times, rather, it indicated periods of crepuscular movement in both upstream and downstream directions. The multi-antenna reader set up at Site 2 in 2020 recorded 5140 total detections. Unfortunately, due to a malfunction of the antenna reader’s multiplexer, the antenna number recorded for each detection was found to contain frequent errors, so I am not able to reliably report on migration direction at this site. Site 3 displayed clear and easily discernable migration events, likely due, in part, to the favorable 81 configuration of the antennae in relation to the thermal refuge. At this site, fish moved through a narrow channel between the site and the mainstem of the river and did not tend to make multiple upstream-downstream movements in a single migration event, as was observed at Site 1. From 713 total detection records, I identified 280 migration events, 145 emigrations and 135 immigrations, between 2020 and 2021. Figure 3.6 displays diel migration frequency, parsed into immigration and emigration events, for each available site. Site 4 recorded the fewest total detections (n = 76) and the fewest migration events (n= 34), even though I captured the most fish at this site (n = 50). I identified only 22 emigrations and 12 immigrations, indicating that most fish reared in this thermal refuge continuously and, if they did emigrate, were not likely to return. Figure 3.6. Diel migration frequency by Site. Note that migration direction could not be interpreted for Site 2/Site 2/2S. 82 Based on the location of modes in the diel frequency of detections and migrations, it appeared that fish were engaging in crepuscular movements, and were generally occupying mainstem habitats at night and thermal refuges during the day, as was most clearly indicated at Site 3. To further visualize the effects of photoperiod on fish movement and habitat occupancy, I plotted the migration times for Site 3 and the bulk detection times for Sites 1, 2, 2S (indicating site occupancy), and 4 against sunrise and sunset times over the data collection periods. This data appears to indicate that fish move between thermal refuges and mainstem habitats in response to light conditions, as the time between movement events corresponds to and decreases with shortening day length (Figure 3.7). My detection data indicates that fish occupied thermal refuges in the summer during periods of high mainstem temperatures in the daytime. However, fish were also found to occupy thermal refuges across a range of mainstem river temperatures, even into the fall when mainstem temperatures crossed over and became cooler than thermal refuge temperatures. 83 Figure 3.7. Emigration (E), Immigration (I), and Detection (D) times associated with Sunrise (SR) and Sunset (SS) events at all sites. 84 DHM Modelling My occupancy model was limited to fish that made multi-day, round-trip migrations at thermal refuges, as occupancy could only be inferred up to each fish’s penultimate migration event. In the case of my data, this limited me to nine fish and 264 migration events, all at Site 3. With occupancy logged at time t in 15-minute intervals, concurrent with mainstem and thermal refuge temperatures, my occupancy data included 14,016 total observations. In modelling the effects of Time, Mainstem Temperature, and Thermal Refuge Temperature on Thermal Refuge Occupancy, the effects of Thermal Refuge Temperature were found to not be significant (P = 0.492), undoubtedly due, in part, to the extremely low variability in thermal refuge temperature at Site 3 (mean = 9.09 °C, SD = 0.16 °C in 2020; mean = 10.15 °C, SD = 0.48 °C in 2021). Between my two models: individual smoother terms for predictor variables Time and Mainstem Temperature (Figure 3.8), and a global smoother of interaction between Time and Mainstem Temperature (Figure 3.9) - both models performed equally well in predicting thermal refuge use in untested data using AUC (AUC = 0.9974 and 0.9972, respectively). My modelling shows that, although the effects of mainstem temperatures on thermal refuge occupancy are significant, they only slightly increase the probability of occupancy at temperature extrema (i.e., above 20°C and below 10°C) (Figure 3.8). Time of day produced the strongest effects on thermal refuge occupancy as fish generally occupied thermal refuges between sunrise and sunset. Moreover, the probability of thermal refuge occupancy is highest even as mainstem temperature continues to decrease towards mid-morning. My model using a global smoother of the interaction effects of time and mainstem temperature on thermal refuge occupancy (Figure 3.9) likely resulted in under-fitting of my Time model term and over-fitting my Mainstem Temperature model term. 85 Figure 3.8. Partial effects plots of individual smoother terms on thermal refuge occupancy with 95% confidence intervals. Row 1 displays partial effects on log-odds scale. Row 2 has been converted to probability scale and incorporates model intercept and uncertainty. y-axis labels show effective degrees of freedom of smoother term, indicating degree of wiggliness. 86 Figure 3.9. Partial effects plot of thermal refuge occupancy displaying interaction between model terms Time and Mainstem Temperature. Points indicate discrete occupancy records across Time and Mainstem Temperature used in the model. Concurrent with my logistic GAMM, I modelled the effects of Time and Mainstem Temperature on DHM at Site 3 to further investigate the effects that each model term contributes to fish movement (vs. occupancy). My model with individual smoothers for each term (Figure 3.10) scored better under AIC than my model with a global smoother of interaction effects between Time and Mainstem Temperature (Figure 3.11) (AIC scores = 660.94 and 753.26, respectively). Both models show pronounced positive partial effects associated with crepuscular time periods and mainstem temperatures between 15-20°C. 87 Figure 3.10. Partial effects plots of individual smoother terms on DHM detections at Site 3 with 95% confidence intervals. y-axis labels show effective degrees of freedom of smoother term, indicating degree of wiggliness. Figure 3.11. Partial effects plot of DHM detections at Site 3 displaying interaction between model terms Time and Mainstem Temperature. 88 Discussion Stream temperature differences between thermal refuges and adjacent mainstem habitats varied by site and throughout the data collection period. All sites displayed cooler thermal refuge temperatures during the summer; however, with the exception of Site 3, average differences in maximum diel temperatures were not drastically cooler, being only between 0.4 °C and 2.3 °C across Sites 1, 2, and 4 during the 2020 and 2021 data collection periods. My observations of consistent use of these habitats highlights the importance that even minor differences in thermal environments have on salmonid life histories and individual behaviours. Early work on salmonid thermal tolerance, as summarized by McCullough et al. (2001), would suggest that fish at these sites were consistently exposed to lethal and sublethal temperature conditions that would have impaired feeding and metabolism. More recent work has suggested, however, that ecosystem productivity, in terms of food availability for fish, helps to mitigate the adverse effects of high stream temperatures on juvenile salmon growth and survival (Brewitt et al. 2017; Lusardi et al. 2020). I found that juvenile salmonids of my focal species made diel migrations between thermal refuge and mainstem habitats at dawn and dusk across a range of mainstem river temperatures. These migrations generally allowed fish to utilize habitats that were cooler during the day, in the summer months, and warmer during the day in the fall; DHMs presumably improving their growth and survival by moving towards temperature optimums, and away from the extremes experienced in mainstem habitats. The cross-over of maximum diel stream temperatures between thermal refuges and mainstem habitats from summer to fall was observed for all sites except Site 1 (Clapperton Creek), likely indicating the influence of groundwater on temperature at Sites 2, 3, and 4. Temperature, however, does not appear to be the primary driver 89 of the timing of DHM, as I expected, and thermal refuge occupancy is strongly linked to photoperiod. These assertions do not imply that stream temperature is unimportant to juvenile salmonids, but that their movement ecology is complex and cannot be reduced to temperature alone; an environmental factor that dominates the literature on the subject and has potentially resulted in over-simplification of solutions for temperature-sensitive streams. The implications of these finding highlight the potential sensitivity of juvenile salmonids to changes in thermal regimes, as their plasticity in response to river and body temperature may be limited. These findings also highlight specific complexities with respect to environmental cues to movement in ectothermic organisms. Differences in frequency of DHM between my sites also suggest that thermal refuge use may be linked to resource availability within thermal refuges. I captured and tagged the most fish at Site 1 and Site 4, and it was these sites that returned the fewest detections per tagged fish. Unlike Sites 2/2S and Site 3, these sites were larger and, perhaps, provided enough food resources to limit density-dependent interactions. Site 1 was observed to have abundant invertebrate life as it was in a downstream position to an extensive trophic cascade. Site 4, although off-channel only, had a substantial pond at its upstream end with abundant primary productivity to fuel invertebrate production for fish. Conversely, Sites 2/2S and Site 3 were limited in size and potential productivity, which may have contributed to my observations of more frequent and regular DHM patterns at these sites. These patterns of juvenile salmon movement in association with stream temperatures and food resources have been welldocumented by others (Armstrong et al. 2013; Armstrong and Schindler 2013; Brewitt and Danner 2014; Baldock et al. 2016; Brewitt et al. 2017), and undoubtedly influenced the patterns I observed. Specifically, Brewitt et al. (2017) found that density-dependent food limitations in 90 food-limited thermal refuges resulted in juvenile Pacific salmon foraging on mainstem prey while using thermal refuges to meet their physiological requirements. My modelling of DHM at Site 3 allowed for an almost idealized scenario in understanding the effects of stream temperature and photoperiod on the timing of juvenile salmon movement, as I had extreme differences in diel temperature maxima and variability between thermal refuges and mainstem habitats at this site. In terms of my modelling, mainstem temperature had only a minor influence on movement patterns compared to photoperiod, and thermal refuge temperature was not significant. Fish at this site made regular, diel migrations between thermal refuge and mainstem habitats at dawn and dusk; however, as noted previously, space and trophic resource appeared to be limited at this site in comparison to Sites 1 and 4 where fish spent prolonged periods of time and were infrequently detected. Looking at the regularity of DHM among individual fish that I observed led me to three possible assumptions: (1) that daily emigrations to mainstem habitats, likely associated with feeding forays (Armstrong et al. 2103; Brewitt et al. 2017), may not have resulted in satiation, as gastric evacuation from satiation would typically take longer than 24 hours (Benkwitt et al. 2009; Armstrong et al. 2013), particularly at the low temperatures associated with Site 3; (2) that offchannel habitats, such as these, provide preferred habitat conditions during daylight hours, beyond temperature alone; and, (3) that juvenile salmon develop patterns in DHM which are signaled by light levels and that thermal refuge site fidelity can be high, as has been observed by others (Minakawa and Kraft 2005). Appendix E includes an example of a 70 mm FL Coho Salmon captured at Site 3. This fish was observed to make diel migrations between mainstem habitats and thermal refuge from the date tagged (July 30, 2021) to the end of the data collection period (September 17, 2021). This example highlights the regularity in patterns of DHM that can 91 be exhibited by juvenile salmonids despite variability in environmental conditions. However, variability in movement patterns among my entire sample population was high, consistent with other studies in the Nicola Watershed that show high variability in patterns of downstream, presmolt migrations (Shrimpton et al. 2014). My study highlights the importance of thermal and physical habitat heterogeneity for stream-rearing salmonids as fish exploit this heterogeneity daily, throughout changing seasons and conditions. Unfortunately, anthropogenic efforts to constrain streams to single-thread channels continues to result in the loss of existing thermal refuges and an inability to recruit new ones through fluvial processes and natural disturbance regimes (Blanton and Marcus 2013). This homogenization of stream habitat can have population-level effects on the intraspecific diversity in salmon populations, affecting local adaptations to spatial and temporal thermal heterogeneity (Ruff et al. 2011). Recent flood events in the Nicola Watershed (i.e., 2017, 2018 and 2021) have resulted in extensive efforts to further constrain the Nicola River (and tributaries) through bank armouring and diking. These alterations reduce critical freshwater habitat for my focal species during juvenile life-stages that have been shown to strongly determine lifetime salmon productivity in the Nicola River Watershed (Warkentin et al. 2022). Homogenization of diel thermal regimes in the summer is also a concern where thermal refuges occur downstream of small surface-release dams, like the one on Nicola Lake (Figure 3.1). Outflow from these dams is typically associated with higher temperature minima and there is often a cooling pattern associated with increased downstream distance from the point of discharge (Zaidel et al. 2021), which is consistent with my observations of summer temperature minima between sites progressively downstream of the Nicola Lake dam. With my results providing evidence that juvenile salmonids are most active in mainstem habitats at night, 92 nighttime temperature minima become an important metric in their conservation and management. Moreover, effects to nighttime temperature minima should be carefully considered when augmenting stream flows from the epilimnion of storage reservoirs. Climate change is worsening and providing greater uncertainty to the situation by affecting weather events and landscape-level processes, including heat-domes, fires, floods, and droughts. Implementing resilience-based plans for managing salmon habitat (Carlson et al. 2015), therefore, is imperative and needs to include strategic partnerships between governments (Indigenous and non-Indigenous), landowners, and the public to promote the ecological processes and conditions that create thermal heterogeneity and sources of cool water in streams. These processes include, but are not limited to, effective stream discharge (Woltemade and Hawkins 2016), forest-covered watershed (Winkler et al. 2017), healthy riparian areas (Ebersole et al. 2003a; Rayne et al. 2008; Dugdale et al. 2018), lateral channel migration (Hall et al. 2007), and connectivity of streams and floodplains (Wade et al. 2013). 93 Epilogue My research reveals that thermal refuges for salmonids in unconfined alluvial stream channels are spatiotemporally dynamic in their occurrence and composition. Changes to the composition of thermal refuge types following a major channel-forming flood event highlight the importance of fluvial geomorphic processes on unconfined floodplains in the recruitment and maintenance of these habitats. Where they occur, thermal refuges are variable in temperature conditions and sensitivity to environmental covariates: air temperature and stream discharge. Off-channel thermal refuges with strong groundwater upwelling were least sensitive to the heating effects of air temperature but, at some sites, showed sensitivity to rapid changes in stream discharge from dam releases during summer baseflow periods. These pulses of flow, during what is typically a slow and steady decline in local hydrographs, caused the vertical hydraulic gradient at some thermal refuges to decrease until flows returned to baseline. This pattern was also exhibited during the onset of precipitation-generated flow pulses in the fall, after mainstem stream temperatures had cooled. Flow conditions had significant effects on thermal refuge and groundwater temperatures, with variable effects on thermal refuge temperatures, but generally exhibited a cooling effect on groundwater temperatures associated with increases in stream discharge. Mainstem stream temperatures during the summer often exceeded thermal tolerance thresholds for our target species and life stages and, not surprisingly, fish were found to use thermal refuges during those times. Although juvenile salmon and steelhead occupied thermal refuges during peak daytime temperatures, I found that mainstem and thermal refuge temperatures were not significant in triggering diel horizontal migration (DHM) or occupancy at thermal refuges. Instead, fish responded to changing light conditions and would move from 94 mainstem habitats to thermal refuges at sunrise, remaining there throughout the day, and would return to mainstem habitats at sunset. Applications of my research in fisheries and ecosystem management should focus on natural processes that maintain ecosystem resilience to climate change and, where acceptable, restoration of anthropogenically-impacted fluvial and hydrological processes. Phenotypic plasticity in salmonid life history underpins much of our hope in conserving these species in the Anthropocene, and there is strong evidence of plasticity in individual movement ecology in Nicola Watershed salmonids related to both adult and juvenile life history requisites (Shrimpton et al. 2014; Turcotte and Shrimpton 2021). Anthropogenic alterations to salmon habitat, however, have been shown to result in rapid genetic changes in populations, due to either or a combination of phenotypic plasticity or rapid genetic evolution resulting from strong selective pressures (Thompson et al. 2019; Jensen et al. 2022). Salmon populations have been shown to evolve over short time periods in response to anthropogenic changes to freshwater ecosystems, however, these evolutionary responses can be to the detriment of a population’s life history diversity and future resilience to environmental changes (Sturrock 2020; Jensen et al. 2022). They can also limit recovery potential of populations following restoration efforts if alleles for pre-disturbance life history traits are lost or seriously diminished prior to meaningful recovery efforts (Thompson et al. 2019). A relevant example of this can be seen in the decline, or altogether disappearance, of populations of spring-run Chinook salmon in much of the Columbia River Basin (NOAA 2022). There, it remains to be seen whether the spring-run allele occurs frequently enough in remaining fall-run Chinook to re-establish these populations following major restoration efforts like dam removal. I suggest that we may one day be asking similar questions of phenotypic plasticity in relation to juvenile freshwater residency in anadromous Nicola River salmon and steelhead, and 95 those of similar semi-arid stream environments if stream thermal environments continue to degrade. Semi-arid streams with stable flow and temperature conditions have been shown to positively affect freshwater residency periods in juvenile Chinook salmon at individual and population scales, improving growth and body size during outmigration from streams (Roddam and Ward 2017). Conversely, decreased summer baseflows during juvenile rearing, and subsequent increases in stream temperature, have already been limiting Chinook productivity in the Nicola River (Warkentin et al. 2022). Behavioural thermoregulation in poikilotherms represents inherent plasticity that provides resilience to changes in stream thermal regimes, given that there are favorable thermal environments to move to. These movements incur trade-offs, however, as temperature is not the only factor determining habitat suitability (Brönmark et al. 2008; Roy et al. 2013; Kurylyk et al. 2015a; Lo et al. 2022). In addition to this, my research reveals that behavioural plasticity in thermoregulation may also be trumped by external factors that are not stochastic – namely, photoperiod. Other studies have shown DHM of juvenile salmonids in response to photoperiod, particularly in downstream smolt migrations at night (Riley et al. 2012; Riley et al. 2014; Zydlewski et al. 2014; Yamada et al. 2022); however, this is the first study to link round-trip DHMs of juvenile salmonids between mainstem and thermal refuges to photoperiod. My research also presents evidence that behavioural thermoregulation in these fish may be limited by a stronger migration response to sunrise and sunset events. In related studies, juvenile salmon and steelhead have been shown to use thermal refuges for growth and metabolism, while engaging in feeding forays to mainstem habitats that were thermally less optimal (Armstrong et al. 2013; Brewitt et al. 2017). In my study, I show that juvenile salmon and steelhead engage in emigration from thermal refuges to mainstem habitats in response to waning light levels at sunset, 96 and I assume that, based on the small size and limited trophic resources of some of my thermal refuges, these emigrations were for feeding. Conversely, juvenile Chinook salmon in the Columbia River have been found to be most active and to feed during the day and are generally inactive at night (Tiffan et al. 2010). These differences highlight the importance of understanding local variations in ecology at small geographies and point to a greater need for place-based fisheries management (Gayeski et al. 2018). In keeping with the need to implement place-based solutions in the conservation of thermal refuges for juvenile Pacific salmon and steelhead, restoration efforts must also be process-based (Beechie et al. 2010). I support this statement with at least three key findings in my research: that the cut and fill alluviation of lateral stream channel movements during flood events creates new, accessible thermal refuges that are off-channel to mainstem habitat; that these off-channel thermal refuges are least sensitive to thermal degradation of high air temperature events; and, that flow manipulations from dams may have unintended, and potentially negative, consequences on hyporheic exchange processes at thermal refuges. These observations support restoration objectives within the more modern stream evolution model proposed by Cluer and Thorne (2014). Under this model, channel evolution is cyclical, degrading and aggrading, away and towards, respectively, a pre-disturbance form that, in many streams, has not been seen since European colonization of western North America. This pre-disturbance channel form is that of a multi-thread, wetland stream corridor where the invert elevation of the stream channel is close to that of the surrounding floodplain. Under this condition, groundwater and surface water elevations are similar, facilitating hyporheic exchange processes across a frequently wetted floodplain. This model also works well within the concept of shifting habitat mosaics (SHM) (Chapter 1) which had been introduced a decade earlier (Hauer et al. 2003; 97 Hauer and Lorang 2004). SHM becomes compromised anytime anthropogenic alterations limit an ecosystem’s ability to express its natural variability across a floodplain. Examples of this, specific to my study reach, include extensive clearing of riparian forests, armouring of stream banks to counteract subsequent increases in erosion following riparian clearing, and changes to hydrologic regimes that affect a stream’s ability to move laterally and create new habitats (Chapter 1) and to exchange water vertically between surface water and groundwater (Chapter 2). It is the natural variability in ecosystems that produces heterogeneity in stream physical and thermal environments which my research provides further evidence of being important to the freshwater life history of juvenile Pacific salmon and steelhead (Chapter 3). My dissertation investigates three important components of stream thermal refuges for juvenile Pacific salmon and steelhead: their composition and spatiotemporal dynamics; the in situ environmental conditions and variables at these sites; and, associated fish behaviour and habitat use. I have unified these components spatially within a single study reach, however, each phase (as represented by each chapter) stands alone temporally as a natural progression of my research, where the preceding work informs what follows. In future research, I would like to unify thermal refuge spatiotemporal dynamics, all in situ conditions (including vertical groundwater fluxes), and associated fish behaviour by monitoring concurrently. I also recognize a need to better understand how the size of thermal refuges and availability of food within them affects DHM patterns of the fish that use them, and how factors outside of the channel affect trophic resources within. As informed by my results, I have raised concerns about the effects of flow manipulations from Nicola Lake dam on environmental conditions at thermal refuges. Where appropriate, I would like to investigate this further through experimental manipulation during critical rearing periods. This will help researchers and practitioners better understand the 98 causal mechanisms of groundwater and temperature dynamics at thermal refuges in relation to stream discharge and will help inform future decision-making related to environmental flows. Finally, thermal refuges in streams where ambient temperatures exceed critical thresholds for survival, like the Nicola River, provide opportunities for improved survival of juvenile Pacific salmon and steelhead. An important next step will be to quantify the effects of thermal refuges on population productivity under future climate and floodplain habitat scenarios. 99 References Achord S, Matthews G, Johnson O, Marsh D. 1996. Use of passive integrated transponder (PIT) tags to monitor migration timing of Snake River Chinook salmon smolts. North Am J Fish Manage. 16: 302-313. Acolas ML, Roussel JM, Lebel JM, Baglinière JL. 2007. Laboratory experiment on survival, growth and tag retention following PIT injection into the body cavity of juvenile brown trout (Salmo trutta). Fish Res. 86: 280-284. Alekseychik P, Katul G, Korpela I, Launiainen S. 2021. Eddies in motion: visualizing boundarylayer turbulence above an open boreal peatland using UAS thermal videos. Atmos Meas Tech. 14: 3501-3521. Ameijeiras-Alonso J, Crujeiras RM, Rodríguez-Casal A. 2019. Mode testing, critical bandwidth and excess mass. Test 28: 900-919. Ameijeiras-Alonso J, Crujeiras RM, Rodríguez-Casal A. 2021. Multimode: an R package for mode assessment. J Stat Soft. 97: 1-32. Armstrong JB, Schindler DE, Ruff CP, Brooks GT, Bentley KE, Torgersen CE. 2013. Diel horizontal migration in streams: Juvenile fish exploit spatial heterogeneity in thermal and trophic resources. Ecology. 94: 2066–2075. Armstrong JB, Schindler DE. 2013. Going with the flow: spatial distributions of juvenile coho salmon track an annually shifting mosaic of water temperature. Ecosystems. 16: 1429– 1441. 100 Baker E, Lautz L, Kelleher C, McKenzie J. 2018. The importance of incorporating diurnally fluctuating stream discharge in stream temperature energy balance models. Hydrol Process. 32: 2901-2914. Bal G, Rivot E, Prévost E, Piou C, Baglinière JL. 2011. Effect of water temperature and density of juvenile salmonids on growth of young-of-the-year Atlantic salmon Salmo salar. J Fish Biol. 78: 1002-1022. Baldock JR., Armstrong JB, Schindler DE, Carter JC. 2016. Juvenile coho salmon track a seasonally shifting thermal mosaic across a river floodplain. Freshw Biol. 61: 1454-1465. Baxter CV, Hauer FR, Woessner WW. 2003. Measuring groundwater-stream water exchange: new techniques for installing minipiezometers and estimating hydraulic conductivity. Trans Am Fish Soc. 132: 493-502. Baxter J, McPhail JD. 1999. The influence of redd site selection, groundwater upwelling, and over-winter incubation temperature on survival of bull trout (Salvelinus confluentus) from egg to alevin. Can J Zool. 77: 1233-1239. Beechie TJ, Sear DA, Olden JD, Pess GR, Buffington JM, Moir H, Roni P, Pollock MM. 2010. Process-based principles for restoring river ecosystems. BioScience. 60(3): 209-222. Benkwitt CE, Brodeur RD, Hurst TP, Daly EA. 2009. Diel feeding chronology, gastric evacuation, and daily food consumption of juvenile Chinook salmon in Oregon coastal waters. Trans Am Fish Soc. 138: 111-120. Berghuijs WR, Woods RA, Hrachowitz M. 2014. A precipitation shift from snow towards rain leads to a decrease in streamflow. Nat Clim Change. 4pp. 101 Blanton P, Marcus WA. 2013. Transportation infrastructure, river confinement, and impacts of floodplain and channel habitat, Yakima and Chehalis rivers, Washington, U.S.A. Geomorphology. 189: 55–65. Bray EN, Dozier J, Dunne T. 2017. Mechanics of the energy balance in large lowland rivers, and why the bed matters. Geophys Res Lett. 44: 8910-8918. Breau C, Cunjuk RA, Bremset G. 2007. Age-specific aggregation of wild juvenile Atlantic salmon Salmo salar at cool water sources during high temperature events. J Fish Biol. 71: 1179–1191. Breau C, Cunjuk RA, Peake SJ. 2011. Behaviour during elevated water temperature: can physiology explain movement of juvenile Atlantic salmon to cool water? J Anim Ecol. 80: 844–853. Brennan SR, Schindler DE, Cline TJ, Walsworth TE, Buck G, Fernandez DP. 2019. Shifting habitat mosaics and fish production across river basins. Science. 364: 783–786. Brett JR. 1952. Temperature tolerance in young Pacific salmon, genus Oncorhynchus. J Fish Res Board Can. 9(6): 265-323. Brewitt KS, Danner EM. 2014. Spatio-temporal temperature variation influences juvenile steelhead (Oncorhynchus mykiss) use of thermal refuges. Ecosphere. 5(7): 92, 26 pp. Brewitt KS, Danner EM, Moore JW. 2017. Hot eats and cool creeks: juvenile Pacific salmonids use mainstem prey while in thermal refuges. Can J Fish Aquat Sci. 74: 1588–1602. Briggs MA, Goodling P, Johnson ZC, Rogers KM, Hitt NP, Fair JB, Snyder CD. 2022. Bedrock depth influences spatial patterns of summer baseflow, temperature and flow disconnection for mountainous headwater streams. Hydrol Earth Syst Sci. 26: 3989-4011. 102 Brönmark C, Skov C, Brodersen J, Nilsson PA, Hansson L. 2008. Seasonal migration determined by a trade-off between predator avoidance and growth. PLoS ONE. 3(4): e1957, 7 pp. Burke C, Wich S, Kusin K, McAree O, Harrison ME, Ripoll B, Ermiasi Y, Mulero-Pázmány M, Longmore S. 2019. Thermal-drones as a safe and reliable method for detecting subterranean peat fires. Drones. 3(23): 1–16. Burkholder BK, Grant GE, Haggerty R, Khangaonkar T, Wampler PJ. 2008. Influence of hyporheic flow and geomorphology on temperature of a large, gravel-bed river, Clackamas River, Oregon, USA. Hydrol Process. 22: 941-953. Caissie D. 2006. The thermal regime of rivers: a review. Freshw Biol. 51:1389–1406. Caissie D, Luce CH. 2017. Quantifying streambed advection and conduction heat fluxes. Water Resour Res. 53: 1595-1624. Campbell EY, Dunham JB, Reeves GH. 2020. Linkages between temperature, macroinvertebrates, and young-of-year Coho Salmon growth in surface-water and groundwater streams. Freshw Sci. 39(3): 447-460. Carlson AK, Taylor WW, Schlee KM, Zorn TG, Infante DM. 2017. Projected impacts of climate change on stream salmonids with implications for resilience-based management. Ecology of Freshw Fish. 26: 190-204. Caruso A, Ridolfi L, Boano F. 2016. Impact of watershed topography on hyporheic exchange. Adv Water Resour. 94(2016): 400-411. Casas-Mulet R, Pander J, Ryu D, Stewardson M, Geist J. 2020. Unmanned aerial vehicle (UAV)based thermal infra-red (TIR) and Optical imagery reveals multi-scale controls of coldwater areas over a groundwater-dominated riverscape. Front Environ Sci. 8(64): 1–16. 103 Cluer B, Thorne C. 2014. A stream evolution model integrating habitat and ecosystem benefits. River Res Appl. 30: 135-154. Conrad JL, Holmes E, Jeffres C, Takata L, Ikemiyagi N, Katz J, Sommer T. 2016. Application of passive integrated transponder technology to juvenile salmon habitat use on an experimental agricultural floodplain. North Am J Fish Manage. 36(1): 30-39. Cristea NC, Burges SJ. 2010. An assessment of the current and future thermal regimes of three streams located in the Wenatchee River basin, Washington State: some implications for regional river basin systems. Clim Change. 102: 493-520. Crozier LG, Burke BJ, Chasco BE, Widener DL, Zabel RW. 2021. Climate change threatens Chinook salmon throughout their life cycle. Commun Biol. 4:222, 14 pp. Cunningham DS, Braun DC, Moore JW, Martens AM. 2023. Forestry influences on salmonid habitat in the North Thompson River watershed, British Columbia. Can J Fish Aquat Sci. 00: 1-18 (2023). Curry RA, Noakes DLG, Morgan GE. 1995. Groundwater and the incubation and emergence of brook trout. Can J Fish Aquat Sci. 52: 1741–1749. David AT, Asarian JE, Lake FK. 2018. Wildfire smoke cools summer river and stream water temperatures. Water Resour Res. 54: 7273-7290. Davidson RS, Letcher BH, Nislow KH. 2010. Drivers of growth variation in juvenile Atlantic salmon (Salmo salar): an elasticity analysis approach. Anim Ecol. 79: 1113-1121. Dittman AH, Quinn TP. 1996. Homing in Pacific salmon: mechanisms and ecological basis. J Exp Biol. 199: 83–91. Dobson D, Holt K, Davis B. 2020. A technical review of the management approach for stream-type Fraser River Chinook. DFO Can. Sci. Advis. Sec. Res. Doc. 2020/027. 280pp. 104 Dugdale SJ, Bergeron NE, St-Hilaire A. 2013. Temporal variability of thermal refuges and water temperature patterns in an Atlantic salmon river. Remote Sens Environ. 136: 358–373. Dugdale SJ, Bergeron NE, St-Hilaire A. 2015. Spatial distribution of thermal refuges analysed in relation to riverscape hydromorphology using airborne thermal infrared imagery. Remote Sens Environ. 160: 43–55. Dugdale SJ, Kelleher CA, Malcolm IA, Caldwell S, Hannah DM. 2019. Assessing the potential of drone-based thermal infrared imagery for quantifying river temperature heterogeneity. Hydrol Process. 33: 1152–1163. Dugdale SJ, Malcolm IA, Kantola KK, Hannah DM. 2018. Stream temperature under contrasting riparian forest cover: understanding thermal dynamics and heat exchange processes. Sci Total Environ. 610-611: 1375-1389. Dunham J, Chandler G, Rieman B, Martin D. 2005. Measuring stream temperature with digital data loggers: a user guide. Gen. Tech. Rep. RMRS-GTR150WWW. Fort Collins, CO: U.S. Department of Agriculture, Forest Service, Rocky Mountain Research Station. 15 pp. Ebersole JL, Liss WJ, Frissel CA. 2001. Relationship between stream temperature, thermal refugia and rainbow trout Oncorhynchus mykiss abundance in arid-land streams in the northwestern United States. Ecol Freshw Fish. 10:1–10. Ebersole JL, Liss WJ, Frissel CA. 2003a. Cold water patches in warm streams: physiochemical characteristics and the influence of shading. J Am Water Resour Assoc. 39(2): 355–368. Ebersole JL, Liss WJ, Frissel CA. 2003b. Thermal heterogeneity, stream channel morphology, and salmonid abundance in northeastern Oregon streams. Can J Fish Aquat Sci. 60: 1266–1280. 105 Ebersole JL, Quiñones RM, Clements S, Letcher BH. 2020. Managing climate refugia for freshwater fishes under an expanding human footprint. Front Ecol Environ. 18(5): 271280. Farrell AP, Hinch SG, Cooke SJ, Patterson DA, Crossin GT, Lapointe M, Mathes MT. 2008. Pacific salmon in hot water: applying aerobic scope models and biotelemetry to predict the success of spawning migrations. Physiol Biochem Zool. 81(6): 697-708. Fullerton AH, Torgersen CE, Lawler JJ, Steel EA, Ebersole JL, Lee SY. 2018. Longitudinal thermal heterogeneity in rivers and refugia for coldwater species: effects of scale and climate change. Aquat Sci. 80(3): 15pp. Gayeski NJ, Stanford JA, Montgomery DR, Lichatowich J, Peterman RM, Williams RN. 2018. The failure of wild salmon management: need for a place-based conceptual foundation. Fisheries. 43(7): 303-309. Getis A, Ord JK. 1992. The analysis of spatial association by use of distance statistics. Geog Anal. 24(3): 18pp. Government of Canada. 2022. Species at risk public registry. Available from: https://www.canada.ca/en/environment-climate-change/services/species-risk-publicregistry.html Gries G, Letcher BH. 2002. Tag retention and survival of age-0 Atlantic salmon following surgical implantation with passive integrated transponder tags. North Am J Fish Manage. 22: 219–222. Guenther SM, Gomi T, Moore RD. 2014. Stream and bed temperature variability in a coastal headwater catchment: influences of surface-subsurface interactions and partial-retention forest harvesting. Hydrol Process. 28: 1238-1249. 106 Hall JE, Holzer DM, Beechie TJ. 2007. Predicting river floodplain and lateral channel migration for salmon habitat conservation. J Am Water Resour Assoc. 43(3): 786-797. Harris FC, Peterson EW. 2020. 1-D vertical flux dynamics in a low-gradient stream: an assessment of stage as a control of vertical hyporheic exchange. Water. 12(708): 15 pp. Hasler CT, Guimond E, Mossop B, Hinch SG, Cooke SJ. 2014. Effectiveness of pulse flows in a regulated river for inducing upstream movement of an imperiled stock of Chinook salmon. Aquat Sci. 76: 231-241. Hauer FR, Dahm CN, Lamberti GA, Stanford JA. 2003. Landscapes and ecological variability of rivers in North America: factors affecting restoration strategies. In: Wissmar RC, Bisson PA, editors. Strategies for restoring river ecosystems: sources of variability and uncertainty in natural and managed systems. Bethesda, Maryland, USA. American Fisheries Society. p. 81–105. Hauer FR, Lorang MS. 2004. River regulation, decline of ecological resources, and potential for restoration in a semi-arid lands river in the western USA. Aquat Sci. 66: 388–401. Hebert C, Caissie D, Satish MG, El-Jabi N. 2011. Study of stream temperature dynamics and corresponding heat fluxes within Miramichi River catchments (New Brunswick, Canada). Hydrol Process. 25: 2439-2455. Hill DJ, Pypker TG, Church JS. 2019. Applications of unpiloted aerial vehicles (UAVs) in forest hydrology. In DF Levia DF, D Carlyle-Moses, S Iida, B Michalzik, K Nanko, A Tischer (eds.) Forest-Water Interactions, Ecological Studies Series No. 240 (pp. 55-85). Switzerland AG: Springer Nature. https://doi.org/10.1007/978-3-030-26086-6_3 Hitt N, Snook E, Massie D. 2017. Brook trout use of thermal refugia and foraging habitat influenced by brown trout. Can J Fish Aquat Sci. 74(3): 406-418. 107 Hyun Y, Kim H, Lee S, Lee K. 2011. Characterizing streambed water fluxes using temperature and head data on multiple spatial scales in Munsan stream, South Korea. J Hydrol. 402: 377-387. Isaak DJ, Wollrab S, Horan D, Chandler G. 2012. Climate change effects on stream and river temperatures across the northwest U.S. from 1980-2009 and implications for salmonid fishes. Clim Change. 113: 499-524. Jensen AJ, Hagen IJ, Czorlich Y, Bolstad GH, Bremset G, Finstad B, Hindar K, Skaal Ø, Karlsson S. 2022. Large-effect loci mediate rapid adaptation of salmon body size after river regulation. Evolution. 119(44): e2207634119, 8 pp. Jensen AM, McKee M, Chen Y. 2014. Procedures for processing thermal images using low-cost microbolometer cameras for small unmanned aerial systems. IEEE Geosci Remote Sens Symp. pp. 2629-2632. Jensen AM, Neilson BT, McKee M, YangQuan C. 2012. Thermal remote sensing with an autonomous unmanned aerial remote sensing platform for surface stream temperatures. IEEE Geosci Remote Sens Symp. pp. 5049–5052. Käser DH, Binley A, Heathwaite AL, Krause S. 2009. Spatio-temporal variations of hyporheic flow in a riffle-step-pool sequence. Hydrol Process. 23: 2138-2149. Kelly J, Kjlun N, Olsson P, Mihai L, Liljeblad B, Weslien P, Klemedtsson L, Eklundh L. 2019. Challenges and best practices for deriving temperature data from an uncalibrated UAV Thermal Infrared Camera. Remote Sens. 11(567): 21 pp. Keefer ML, Clabough TS, Jepson MA, Johnson EL, Peery CA, Caudill CC. 2018. Thermal exposure of adult Chinook salmon and steelhead: diverse behavioural strategies in a large and warming river system. PLoS ONE. 13(9): e0204274, 30 pp. 108 Kleindl WJ, Rains MC, Marshall LA, Hauer FR. 2015. Fire and flood expand the floodplain shifting habitat mosaic concept. Fire Ecol. 34: 1366-1382. Kosakoski GT, Hamilton RE. 1982. Water requirements for the fisheries resource of the Nicola River, B.C. Can Manuscr Rep Fish Aquat Sci. 1680: 127 pp. Krause S, Blume T, Cassidy NJ. 2012. Investigating patterns and controls of groundwater upwelling in a lowland river by combining fibre-optic distributed temperature sensing with observations of vertical hydraulic gradients. Hydrol Earth Syst Sci. 16: 1775-1792. Kurylyk BL, MacQuarrie KTB, Voss CI. 2014. Climate change impacts on the temperature and magnitude of groundwater discharge from shallow, unconfined aquifers. Water Resour Res. 50: 3253-3274. Kurylyk BL, MacQuarrie KTB, Linnansaari T, Cunjak RA, Curry RA. 2015a. Preserving, augmenting, and creating cold-water thermal refugia: concepts derived from research on the Miramichi River, New Brunswick (Canada). Ecohydrology. 8: 1095–1108. Kurylyk BL, MacQuarrie KTB, Caissie D, McKenzie JM. 2015b. Shallow groundwater thermal sensitivity to climate change and land cover disturbances: derivation of analytical expressions and implications for stream temperature modelling. Hydrol Earth Syst Sci. 19: 2469-2489. Lewis D. 2016. Current condition and 10-year historic trend analysis of hydrologic hazards in the Merritt Timber Supply Area. Ministry of Forests, Lands and Natural Resource Operations. 31pp. Lloyd D, Angrove K, Hope G, Thompson C. 1990. A guide to site identification and interpretation for the Kamloops Forest Region. Research Branch, Ministry of Forests, Province of British Columbia. Victoria, B.C. 109 Lo VK, Martin BT, Danner EM, Cocherell DE, Cech JJ, Fangue NA. 2022. The effect of temperature on specific dynamic action of juvenile fall-run Chinook salmon, Oncorhynchus tshawytscha. Conserv Physiol. 10(1): coac067, 12pp. Ludovisi R, Tauro F, Salvati R, Khoury S, Mugnozza GS, Harfouche A. 2017. UAV-based thermal imaging for high-throughput field phenotyping of black poplar response to drought. Front Plant Sci. 8(1681): 18 pp. Lusardi RA, Hammock BG, Jeffries CA, Dahlgren RA, Kiernan JA. 2020. Oversummer growth and survival of juvenile coho salmon (Oncorhynchus kisutch) across a natural gradient of stream water temperature and prey availability: an in situ enclosure experiment. Can J Fish Aquat Sci. 77: 413-424. Mantua N, Tohver I, Hamlet A. 2010. Climate change impacts on streamflow extremes and summertime stream temperature and their possible consequences for freshwater salmon habitat in Washington State. Clim Change 102: 187-223. Marzadri A, Tonina D, McKean JA, Tiedemann MG, Benjankar R. 2014. Multi-scale streambed topographic and discharge effects on hyporheic exchange at the stream network scale in confined streams. J Hydrol. 519: 1997–2011. Mauger S, Shaftel R, Leppi JC, Rinella DJ. 2017. Summer temperature regimes in southcentral Alaska streams: watershed drivers of variation and potential implications for Pacific salmon. Can J Fish Aquat Sci. 74: 702-715. Mathews M, Bocking B, Glova G, Todd N, Sampson T. 2007. Development of an annual salmonid productivity assessment program for the Nicola River Watershed. Nicola Tribal Association - Nicola Watershed Stewardship and Fisheries Authority. Available from: https://www.psf.ca/sites/default/files/FSWP_07_D2_NTA_FINAL_REPORT.pdf 110 McCallum AM, Andersen MS, Giambastiani BMS, Kelly BFJ, Acworth RI. 2013. River-aquifer interactions in a semi-arid environment stressed by groundwater abstraction. Hydrol Process. 27: 1072-1085. McCallum JL, Shanafield MS. 2016. Residence times of stream-groundwater exchanges due to transient stream stage fluctuations. Water Resour Res. 52: 2059-2073. McCullough DA. 1999. A review and synthesis of the effects of alterations to the water temperature regime on freshwater life stages of salmonids, with special reference to Chinook salmon. U.S. Environmental Protection Agency, Seattle, Washington. McCullough DA, Shelley S, Sturdevant D, Hicks M. 2001. Summary of the technical literature examining the physiological effects of temperature on salmonids. U.S. Environmental Protection Agency. EPA-910-D-01-005. Menichino, GT, Scott D, Hester ET. 2015. Abundance and dimensions of naturally occurring macropores along stream channels and the effects of artificially constructed large macropores on transient storage. Freshw Sci. 34(1): 125-138. Minakawa N, Kraft GF. 2005. Homing behaviour of juvenile coho salmon (Oncorhynchus kisutch) within an off-channel habitat. Ecol Freshw Fish. 14: 197-201. Ministry of Environment (MOE). 2022. Watershed Dictionary Query. British Columbia Ministry of Environment. Available from: http://a100.gov.bc.ca/pub/fidq/viewWatershedDictionary.do Moore RD, Nelitz M, Parkinson E. 2013. Empirical modelling of maximum weekly average stream temperature in British Columbia, Canada, to support assessment of fish habitat suitability. Can Water Resour J. 38(2): 135-147. 111 Murray J, Ayers J, Brookfield A. 2023. The impact of climate change on monthly baseflow trends across Canada. J Hydrol. 618, 2023, 129294: 10 pp. Nathan R, Getz WM, Revilla E, Holyoak M, Kadmon R, Saltz D, Smouse PE. 2008. A movement ecology paradigm for unifying organismal movement research. Proc Natl Acad Sci USA. 105: 19052-19059. Neiffer DL, Stamper MA. 2009. Fish sedation, anesthesia, analgesia, and euthanasia: considerations, methods, and types of drugs. ILAR Journal. 50(4): 343-360. Nielsen, JL, Lilse TE. 1994. Thermally stratified pools and their use by steelhead in Northern California streams. Trans Am Fish Soc. 123: 613-626. National Oceanic and Atmospheric Administration (NOAA). 2022. 2022 5-Year Review: summary and evaluation of Upper Columbia River spring-run Chinook salmon and Upper Columbia River steelhead. National Marine Fisheries Service, West Coast Region. Available from: 2022 5-Year Review: Summary & Evaluation of Upper Columbia River Spring-run Chinook Salmon and Upper Columbia River Steelhead (noaa.gov) Ord JK, Getis A. 1995. Local spatial autocorrelation statistics: distributional issues and an application. Geog Anal. 27(4): 21pp. Pedersen EJ, Miller DL, Simpson GL, Ross N. 2019. Hierarchical generalized additive models in ecology: an introduction with mgcv. PeerJ 7:e6876 DOI 10.7717/peerj.6876 Perry RW, Plumb JM. 2015. Using a laboratory-based growth model to estimate mass- and temperature-dependent growth parameters across populations of juvenile Chinook salmon. Trans Am Fish Soc. 144: 331-336. 112 Peterson ML, Fuller AN, Demko D. 2017. Environmental factors associated with the upstream migration of fall-run Chinook salmon in a regulated river. North Am J Fish Manage. 37: 78-93. Poff NL, Allan JD, Bain MB, Karr JR, Prestegaard KL, Richter BD, Sparks RE, Stromberg JL. 1997. The natural flow regime: a paradigm for river conservation and restoration. BioScience. 47(11): 769-784. Poff NL. 2018. Beyond the natural flow regime? Broadening the hydro-ecological foundation to meet environmental flows challenges in a non-stationary world. Freshw Biol. 63:10111021. Poole GC, Berman CH. 2001. An ecological perspective on in-stream temperature: natural heat dynamics and mechanisms of human-caused thermal degradation. Environ Manage. 27(6): 787-802. Power G, Brown RS, Imhof JG. 1999. Groundwater and fish – insights from northern North America. Hydrol Process. 13: 401-422. Quigley JT, Hinch SG. 2006. Effects of rapid experimental temperature increases on acute physiological stress and behaviour of stream dwelling juvenile Chinook salmon. J Therm Biol. 31: 429-441. Quinn, JM, Wright-Stow AE. 2008. Stream size influences stream temperature impacts and recovery rates after clearfell logging. For Ecol Manage. 256: 2101-2109. R Core Team. 2021. R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. Available from: https://www.R-project.org/. 113 Rayne S, Henderson G, Gill P, Forest K. 2008. Riparian forest harvesting effects on maximum water temperatures in wetland-sourced headwater streams from the Nicola River Watershed, British Columbia, Canada. Water Resour Manage. 22: 565–578. Reed TE, Schindler DE, Hague MJ, Patterson DA, Meir E, Waples RS, Hinch SG. 2011. Time to evolve? Potential evolutionary responses of Fraser River sockeye salmon to climate change and effects on persistence. PLoS ONE. 6(6): e20380, 13 pp. Reeder WJ, Gariglio F, Carnie R, Tang C, Isaak D, Chen Q, Yu Z, McKean JA, Tonia D. 2021. Some (fish might) like it hot: habitat quality and fish growth from past to future climates. Sci Total Environ. 787: 147532, 16 pp. Reese-Hansen L, Nelitz M, Parkinson E. 2012. Designating Temperature Sensitive Streams (TSS) in British Columbia: a discussion paper exploring the science, policy, and climate change considerations associated with a TSS designation procedure. B.C. Ministry of Environment, Fisheries Management Report RD# 123. Victoria, B.C.. 71 pp. Richter A, Kolmes SA. 2005. Maximum temperature limits for Chinook, Coho, and Chum Salmon, and Steelhead Trout in the Pacific Northwest. Rev Fish Sci. 13:23–49. Riley WD, Bendall B, Ives MJ, Edmonds NJ, Maxwell DL. 2012. Street lighting disrupts the diel migratory pattern of wild Atlantic salmon, Salmo salar L., smolts leaving their natal stream. Aquaculture. 330-333(2012): 74-81. Riley WD, Ibbotson AT, Maxwell DL, Davison PI, Beaumont WRC, Ives MJ. 2014. Development of schooling behaviour during the downstream migration of Atlantic salmon Salmo salar smolts in a chalk stream. J Fish Biol. 85: 1042-1059. Robin X, Turck N, Hainard A. 2011. pROC: an open-source package for R and S+ to analyze and compare ROC curves. BMC Bioinf. 12(77): 1-8. 114 Roddam M, Ward DM. 2017. Life-history differences of juvenile Chinook salmon Oncorhynchus tshawystscha across rearing locations in the Shasta River, California. Ecol Freshw Fish. 26: 150-159. Roon DA, Dunham JB, Groom JD. 2021. Shade, light, and stream temperature responses to riparian thinning in second-growth redwood forests of northern California. PLoS One. 16(2): e0246822, 25 pp. Rosenau ML, Angelo M. 2003. Conflicts between people and fish for water: two British Columbia salmon and steelhead rearing streams in need of flows. Pac Fish Res Cons Coun. Vancouver, BC, 97 pp. Roy ML, Roy AG, Grant JWA, Bergeron NE. 2013. Individual variability of wild juvenile Atlantic salmon activity patterns: effect of flow stage, temperature, and habitat use. Can J Fish Aquat Sci. 70: 1082-1091. Ruff CP, Schindler DE, Armstrong JB, Bentley KT, Brooks GT, Holtgrieve GW, McGlauflin MT, Torgersen CE, Seeb JE. 2011. Temperature-associated population diversity in salmon confers benefits to mobile consumers. Ecology. 92(11): 2073-2084. Rutherford JC, Meleason MA, Davies-Colley RJ. 2018. Modelling stream shade: 2. predicting the effects of canopy shape and changes over time. Ecol Eng. 120: 487-496. Ryberg KR, Akyzüz FA, Wiche GJ, Lin W. 2016. Changes in seasonality and timing of peak streamflow in snow and semi-arid climates of the north-central United States, 1910-2012. Hydrol Process. 30: 1208–1218. Scheuerell MD, Schindler DE. 2003. Diel vertical migration by juvenile sockeye salmon: empirical evidence for the antipredation window. Ecology. 84(7): 1713-1720. 115 Schmadel NM, Ward AS, Wondzell SM. 2017. Hydrologic controls on hyporheic exchange in a headwater mountain stream. Water Resour Res. 53: 6260-6278. Sears MW, Angilleta MJ Jr. 2015. Costs and benefits of thermoregulation revisted: both the heterogeneity and spatial structure of temperature drive energetic costs. Am Nat. 185(4): 94-102. Shrimpton JM, Warren KD, Todd NL, McRae CJ, Glova GJ, Telmer KH, Clarke AD. 2014. Freshwater movement patterns by juvenile Pacific salmon Oncorhynchus spp. before they migrate to the ocean: oh the places you’ll go! J Fish Biol. 85: 987–1004. Singha K, Navarre-Sitchler A. 2022. The importance of groundwater in critical zone science. Groundwater. 60(1): 27-34. Simpson, G. 2023. gratia package in r. Version 0.8.1. Available on CRAN: https://rdocumentation.org/packages/gratia/versions/0.8.1 Sturrock AM, Carlson SM, Wikert JD, Heyne T, Nusslé S, Merz JE, Sturrock HJW, Johnson RC. 2020. Unnatural selection of salmon life histories in a modified riverscape. Glob Change Biol. 26: 1235-1247. Sullivan CJ, Vokoun JC, Helton AM, Briggs MA, Kurylyk BL. 2021. An ecohydrological typology for thermal refuges in streams and rivers. Ecohydrology. 14: e2295, 15 pp. Surfleet C, Louen J. 2018. The influence of hyporheic exchange on water temperatures in a headwater stream. Water. 10(1615): 15 pp. Thompson TQ, Bellinger MR, O’Rourke SM, Prince DJ, Stevenson AE, Rodrigues AT, Sloat MR, Speller CF, Yang DY, Butler VL, et al. 2019. Anthropogenic habitat alteration leads to rapid loss of adaptive variation and restoration potential in wild salmon populations. PNAS. 116(1): 177-186. 116 Tiffan KF, Kock TJ, Skalicky JJ. 2010. Diel behaviour of rearing fall Chinook salmon. Northwestern Nat. 91: 342-345. Tiffan KF, Perry RW, Connor WP, Mullins FL, Rabe CD, Nelson DD. 2015. Survival, growth, and tag retention in age-0 Chinook salmon implanted with 8-, 9-, and 12-mm PIT tags. North Am J Fish Manage. 35(4): 845-852. Torgersen CE, Ebersole JE, Keenan DM. 2012. Primer for identifying cold-water refuges to protect and restore thermal diversity in riverine landscapes. Seattle, Washington: United States Environmental Protection Agency, 91 pp. Torgersen CE, Faux RN, McIntosh BA, Poage NJ, Norton DJ. 2001. Airborne thermal remote sensing for water temperature assessment in rivers and streams. Remote Sens Environ. 76: 386–398. Torgersen CE, Price DM, Li HW, McIntosh BA. 1999. Multiscale thermal refugia and stream habitat associations of Chinook salmon in northeastern Oregon. Ecol Appl. 9(1): 301-319. Turcotte LA, Shrimpton JM. 2020. Assessment of spawning site fidelity in interior Fraser River Coho salmon Oncorhynchus kisutch using otolith microchemistry, in British Columbia, Canada. J Fish Biol. 97:1833–1841. Valett HM, Fisher SG, Grimm NB, Camill P. 1994. Vertical hydrologic exchange and ecological stability of a desert stream ecosystem. Ecology. 75(2): 548-560. Wade AA, Beechie TJ, Fleishman E, Mantua NJ, Wu H, Kimball JS, Stoms DM, Stanford JA. 2013. Steelhead vulnerability to climate change in the Pacific Northwest. J Appl Ecol. 50: 1093-1104. 117 Walthers LC, Nener JC. 1997. Continuous water temperature monitoring in the Nicola River, B.C., 1994 implications of high measured temperatures for anadromous salmonids. Can Tech Rep Fish Aquat Sci. 2158: 59 pp. Walthers LC, Nener JC. 2000. Water temperature monitoring in selected Thompson River tributaries, B.C., 1996 implications of measured temperatures for anadromous salmonids. Can Tech Rep Fish Aquat Sci. 2306: 69 pp. Warkentin L. 2020. Regimes of river temperature and flow in an interior watershed, and their implications for Chinook salmon [unpublished master’s thesis]. Simon Fraser University. Warkentin L, Parken CK, Bailey R, Moore JW. 2022. Low summer river flows associated with low productivity of Chinook salmon in a watershed with shifting hydrology. Ecol Solutions Evidence. 3:e12124, 12pp. Water Survey of Canada. 2021. Historical hydrometric data search for Nicola River at Spences Bridge – station number 08LG006, 1911-2019. Available from: https://wateroffice.ec.gc.ca/search/historical_e.html Water Survey of Canada. 2023. Real-time hydrometric data. Available from: https://wateroffice.ec.gc.ca/search/real_time_e.html . Welch C, Cook PG, Harrington GA, Robinson NI. 2013. Propagation of solutes and pressure into aquifers following river stage rise. Water Resour Res. 49: 5246-5259. Winkler R, Spittlehouse D, Boon S. 2017. Streamflow response to clear-cut logging in British Columbia’s Okanagan Plateau. Ecohydrology. 10:e1836, 15pp. Woltemade CJ, Hawkins TW. 2016. Stream temperature impacts because of changes in air temperature, land cover and stream discharge: Navarro River Watershed, California, USA. River Res Appl. 32: 2020-2031. 118 Wondzell SM. 2006. Effect of morphology and discharge on hyporheic exchange flows in two small streams in the Cascade Mountains of Oregon, USA. Hydrol Process. 20: 267-287. Wood SN. 2011. Fast stable restricted maximum likelihood and marginal likelihood estimation of semiparametric generalized linear models. J R Stat Soc. 73(1): 3-36. Wood SN. 2020. gamm4: generalized additive mixed models using ‘mgcv’ and ‘lme4’ for estimation. View on CRAN: https://cran.r-project.org/web/packages/gamm4/index.html Yamada T, Urabe H, Nakamura F. 2022. Diel migration pattern of pink salmon fry in small streams. J Fish Biol. 100 : 1088-1092. Yarnell SM, Viers JH, Mount JF. 2010. Ecology and management of the spring snowmelt recession. BioScience. 60(2): 114-127. Yates D, Galbraith H, Purkey D, Huber-Lee A, Sieber J, West J, Herrod-Julius S, Joyce B. 2008. Climate warming, water storage, and Chinook salmon in California’s Sacramento Valley. Clim Change. 91: 335-350. Yeh TJ, Xiang J, Suribhatla RM, Hsu K, Lee C, Wen J. 2009. River stage tomography: a new approach for characterizing groundwater basins. Water Resour Res. 45, W05409: 14 pp. Zaidel PA, Roy AH, Houle KM, Lambert B, Letcher BH, Nislow KH, Smith C. 2021. Impacts of small dams on stream temperature. Ecol. Indic. 120 (106878): 13 pp. Zhang X, Li H, Deng ZD, Leung LR, Skalski JR, Cooke SJ. 2019. On the variable effects of climate change on Pacific salmon. Ecol Modell. 397:95–106. Zydlewski GB, Stich DS, McCormick SD. 2014. Photoperiod control of downstream movements of Atlantic salmon Salmo salar smolts. J Fish Biol. 85: 1023-1041. 119 To July 31, 2017 August 31, 2017 September 30, 2017 October 12, 2017 July 31, 2017 August 31, 2017 September 30, 2017 October 12, 2017 July 31, 2017 August 31, 2017 September 30, 2017 October 12, 2017 July 31, 2017 August 31, 2017 September 30, 2017 October 12, 2017 July 31, 2017 August 31, 2017 September 30, 2017 October 12, 2017 July 31, 2018 August 31, 2018 September 30, 2018 October 30, 2018 July 31, 2018 From July 14, 2017 August 1, 2017 September 1, 2017 October 1, 2017 July 11, 2017 August 1, 2017 September 1, 2017 October 1, 2017 July 18, 2017 August 1, 2017 September 1, 2017 October 1, 2017 July 12, 2017 August 1, 2017 September 1, 2017 October 1, 2017 July 12, 2017 August 1, 2017 September 1, 2017 October 1, 2017 July 16, 2018 August 1, 2018 September 1, 2018 October 1, 2018 July 14, 2018 Period 2 1 1 1 1 5 5 5 5 4 4 4 4 3 3 3 3 2 2 2 2 1 1 1 1 Site 8.78 5.35 9.28 15.09 16.62 6.67 8.68 14.52 15.38 6.37 9.08 14.71 15.57 3.58 6.98 15.00 15.95 5.66 7.18 9.37 9.47 7.38 10.46 16.43 17.38 Min 9.34 8.08 14.30 19.90 19.95 8.67 13.98 18.99 19.50 8.79 14.68 19.13 19.53 8.10 14.83 19.35 19.87 8.23 11.11 11.80 10.85 10.55 16.04 18.50 19.08 Mean 10.06 11.92 20.23 25.51 23.48 11.43 23.29 23.77 24.26 12.11 23.00 23.77 24.35 13.75 26.56 23.97 23.97 10.26 15.95 14.90 14.04 15.57 22.62 22.33 22.24 Max 0.26 1.67 2.28 2.01 1.71 1.16 3.23 1.98 2.10 1.47 3.19 1.86 2.07 2.17 4.12 1.65 1.93 1.01 1.89 1.26 1.00 1.96 3.15 1.11 1.22 SD Stilling well (thermal refuge) temperature (°C) 8.78 7.98 12.59 17.67 17.09 8.98 10.55 11.92 11.72 10.16 11.72 11.53 11.33 7.48 10.65 16.62 16.33 9.37 10.16 9.77 8.88 7.98 11.04 16.71 17.48 Min 8.91 10.03 15.11 19.92 18.80 9.91 11.97 14.21 12.53 11.13 12.52 12.09 11.95 9.39 15.16 19.15 19.71 10.02 10.70 10.20 9.35 10.85 16.17 18.42 18.98 Mean 9.18 12.59 17.67 21.28 19.85 11.04 13.56 23.20 13.65 12.40 12.79 13.17 14.04 12.59 21.57 21.28 22.72 10.55 11.72 13.27 9.97 15.57 22.72 22.33 22.14 Max Piezometer temperature (°C) 0.13 1.39 1.45 0.96 0.74 0.59 1.05 2.82 0.44 0.58 0.24 0.32 0.71 1.28 2.93 0.91 1.54 0.25 0.15 0.22 0.26 1.72 3.06 1.07 1.17 SD 16.37 6.26 10.08 13.55 17.25 6.41 9.41 14.75 15.53 6.36 9.41 14.89 15.56 6.54 9.71 15.70 16.08 6.36 9.81 16.13 16.08 6.74 9.78 16.68 17.30 Min 21.51 8.26 13.28 19.18 21.75 8.88 14.96 19.25 19.59 8.87 14.96 19.27 19.61 8.99 15.15 19.47 19.96 8.91 15.19 19.53 19.81 9.06 15.16 19.55 19.77 Mean 26.16 10.35 16.94 25.28 26.13 12.63 22.90 23.67 24.15 12.49 22.94 23.79 24.22 12.49 22.90 23.09 24.15 12.39 22.73 22.66 23.83 12.32 21.99 22.18 22.13 Max 2.16 0.74 1.54 2.54 1.90 1.60 3.18 1.87 2.03 1.57 3.17 1.85 2.06 1.47 3.17 1.50 1.93 1.49 3.20 1.22 1.83 1.29 3.24 1.10 1.09 SD 120 Mainstem temperature (°C) Appendix A. Summary of temperature minima, mean, maxima, and standard deviation by site and month. Note that mainstem temperature loggers for Sites 3 and 4 were lost in 2018, however Site 2 provides a suitable representation of mainstem temperatures for these sites (as can be seen in 2017). Site 5 was abandoned for the 2018 season due to it being de-watered by lateral shifting of the mainstem and associated sediment deposition. August 31, 2018 September 30, 2018 October 30, 2018 July 31, 2018 August 31, 2018 September 30, 2018 October 30, 2018 July 31, 2018 August 31, 2018 September 30, 2018 October 30, 2018 August 1, 2018 September 1, 2018 October 1, 2018 July 16, 2018 August 1, 2018 September 1, 2018 October 1, 2018 July 17, 2018 August 1, 2018 September 1, 2018 October 1, 2018 4 4 4 4 3 3 3 3 2 2 2 4.62 9.57 12.50 14.80 4.73 10.16 13.27 16.14 9.18 9.57 9.18 7.62 13.57 18.85 20.94 7.68 13.62 19.01 21.35 9.50 9.91 9.56 10.65 18.52 25.42 25.42 10.75 18.52 25.90 25.90 9.77 10.46 10.16 1.31 1.88 2.89 2.51 1.19 1.80 2.75 2.31 0.09 0.21 0.17 6.17 10.85 13.75 16.05 8.18 12.59 11.92 11.92 9.37 9.57 9.18 8.15 13.39 17.84 18.41 10.15 14.49 14.10 12.74 9.62 9.85 9.39 11.04 15.47 20.42 19.76 12.59 15.76 15.47 14.42 9.87 9.97 9.67 0.98 1.24 1.71 0.75 0.75 0.87 1.14 0.83 0.08 0.12 0.13 - - - - - - - - 4.79 10.54 13.64 - - - - - - - - 7.72 13.70 19.22 - - - - - - - - 10.59 17.99 26.33 121 - - - - - - - - 1.19 1.72 2.65 Appendix B. Plots of time-series mean temperatures by site and year. 122 123 Appendix C. Regressions plots ±SE of river temperature, stilling well (i.e., thermal refuge) temperature, and piezometer (i.e., groundwater) temperature in response to environmental covariates: air temperature, stream discharge, and vertical hydraulic gradient – across all sites and years. 124 125 Appendix D. Time-series daily mean values of stream discharge – Q (m3s-1), stilling well water level – SWWL (m), piezometer water level – PzWL (m), and vertical hydraulic gradient – VHG for 2017 and 2018. 126 Appendix E. Diel migration pattern of a 70 mm Coho Salmon at Site 3 between thermal refuge and mainstem habitats. Date Time Migration 2021-07-30 2021-07-31 2021-07-31 2021-08-01 2021-08-01 2021-08-02 2021-08-02 2021-08-03 2021-08-03 2021-08-04 2021-08-04 2021-08-05 2021-08-05 2021-08-06 2021-08-06 2021-08-07 2021-08-07 2021-08-08 2021-08-08 2021-08-09 2021-08-09 2021-08-10 2021-08-10 2021-08-11 2021-08-11 2021-08-12 2021-08-12 2021-08-13 2021-08-13 2021-08-14 2021-08-14 2021-08-15 2021-08-15 2021-08-16 2021-08-16 2021-08-17 2021-08-17 2021-08-18 2021-08-18 2021-08-19 2021-08-19 2021-08-20 2021-08-20 2021-08-21 21:31:13 04:50:55 21:16:00 04:49:29 21:19:42 05:01:35 21:38:18 04:36:31 21:24:41 04:52:29 21:06:56 04:47:28 20:52:01 04:50:13 20:55:37 04:58:54 21:07:00 05:03:15 21:03:22 05:10:56 20:53:47 05:03:35 20:54:34 04:53:16 20:45:38 05:18:23 20:07:29 05:17:33 20:35:22 05:17:45 19:41:37 05:00:34 20:42:46 05:23:05 20:42:13 05:38:06 20:42:22 05:31:32 20:45:01 05:36:09 20:35:02 05:24:44 20:37:47 06:25:08 Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration 127 Date Time Migration 2021-08-21 2021-08-22 2021-08-22 2021-08-23 2021-08-23 2021-08-24 2021-08-24 2021-08-25 2021-08-25 2021-08-26 2021-08-26 2021-08-27 2021-08-27 2021-08-28 2021-08-28 2021-08-29 2021-08-29 2021-08-30 2021-08-30 2021-08-31 2021-08-31 2021-09-01 2021-09-01 2021-09-02 2021-09-02 2021-09-03 2021-09-03 2021-09-04 2021-09-04 2021-09-05 2021-09-05 2021-09-06 2021-09-06 2021-09-07 2021-09-07 2021-09-08 2021-09-08 2021-09-09 2021-09-09 2021-09-10 2021-09-10 2021-09-11 2021-09-11 2021-09-12 2021-09-12 2021-09-13 20:40:51 05:06:59 20:48:00 04:54:05 20:59:15 05:37:02 20:43:28 05:12:54 21:00:09 05:25:27 20:33:40 05:29:59 20:41:18 05:47:41 20:57:43 05:34:13 20:12:33 05:39:49 20:38:27 05:52:18 20:36:50 05:36:15 20:58:30 05:45:18 20:44:01 05:24:56 20:28:16 05:49:54 20:48:50 05:39:31 20:52:10 05:46:57 20:42:02 05:49:17 20:23:27 06:21:03 20:29:32 06:38:22 20:29:12 06:15:07 20:33:57 06:01:32 20:20:22 06:05:37 20:39:05 06:03:25 Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration 128 Date Time Migration 2021-09-13 2021-09-14 2021-09-14 2021-09-15 2021-09-15 2021-09-16 2021-09-16 2021-09-17 20:27:31 06:20:32 20:25:43 05:36:49 20:20:14 06:06:22 19:58:47 05:55:06 Emigration Immigration Emigration Immigration Emigration Immigration Emigration Immigration 129