THE EFFECT OF PULSED DISCHARGE EVENTS ON THERMAL REFUGIA USE BY BROWN TROUT IN THERMALLY MARGINAL STREAMS A Thesis Presented to the Faculty of the Graduate School of Cornell University In Partial Fulfillment of the Requirements for the Degree of Master of Science by Bethany Ann Boisvert August 2008 © 2008 Bethany Ann Boisvert ABSTRACT Salmonid fishes are known to inhabit streams with ambient summer temperatures approaching or exceeding lethal limits. Under these conditions, localized areas of cool water facilitate the persistence of coldwater fishes, but these may be altered in regulated rivers. In this study we examined the variation in brown trout behavioral thermoregulation within three streams of the Hudson River drainage – the Cedar, Indian and Hudson Rivers – the latter two of which are impacted by recreational discharge events from an upstream dam that supports a summer whitewater rafting industry. We were particularly interested in evaluating the potential of thermal refugia dilution by these flow releases. Based on both laboratory derived tolerance values and field-based thresholds that incorporated metrics of temperature magnitude, duration, and fluctuation, all three rivers were thermally marginal for brown trout during the summers of 2005 and 2006. Behavioral thermoregulation was observed by adult brown trout in our study in all river reaches, albeit infrequently in the Indian River. We found that brown trout in the Cedar River (38%) were more often observed with body temperatures cooler than ambient river temperature than those in either the Hudson (29%) or Indian Rivers (4%). Fewer than 50% of stocked fish persisted over a 67 day period in all three of our study reaches. Persistence of stocked brown trout in the Cedar River in 2006 was greater than in either the Hudson or Indian Rivers in both 2005 and 2006. While recreational discharge events did not alter the mean or maximum daily temperature in either the mainstem Indian or Hudson River reaches, the patches of relatively cool water near tributary confluences were diluted by release events. Both daily temperature maxima and ranges increased significantly at these locations in concert with recreational flow releases. Recreational flow releases were not an important factor accounting for the thermal behavior of brown trout during any time period in the reference Cedar River reach (without dam releases); however, behavioral thermoregulation was reduced during flow releases in both the Indian and Hudson River reaches. In the absence of recreational discharge events, the most important factors affecting behavioral thermoregulation were whether a fish was located near a tributary confluence and the ambient river temperature. Brown trout were consistently cooler relative to ambient river temperature when located near tributaries during times when river temperature was within the upper critical range for brown trout. Behavioral thermoregulation increased as river thermal conditions became more stressful. Our results suggest that accessible thermal refuge areas are important resources that provide brown trout a haven from lethal summer temperature conditions in thermally marginal streams, such as these study reaches in the Upper Hudson River drainage. When low flow conditions correspond with peak summer temperatures, these refuge areas are likely most important and most vulnerable to altered flow regimes. Our results showed that pulsed discharge events altered both the thermal characteristics of refuge areas at tributary confluences and behavioral thermoregulation by stocked brown trout. Although poor survival of these trout in the affected reaches may be due to severe summer temperatures regardless of recreational releases, the observed reduction in behavioral thermoregulation suggests that pulsed discharge events may impair the ability of coldwater fish to survive in regulated systems. BIOGRAPHICAL SKETCH Bethany Ann Boisvert was born in Dover, NH on May 12, 1977 to Paul and Patricia Boisvert. She grew up in the town of Berwick, ME and graduated from Noble High School. She received a Bachelor of Arts Degree in Ecology and Evolutionary Biology from Brown University in May, 1999 where she focused her studies on paleontology and conducted honors research on camel evolution. She discovered the field of fisheries at the Maine Atlantic Salmon Commission, her first job after graduating. Over the following six years she worked in the areas fisheries and climate history at a number of locations. In June of 2005 she began progress towards a Master’s degree in Natural Resources at Cornell University under the advisement of Dr. Clifford Kraft. iii ACKNOWLEDGMENTS I received guidance, support and technical help from many people throughout graduate school. I would like to thank David Post, Edward Camp, Theodore Treska, Geoff Steinhart, Madeleine Mineau, Walt Keller, Beth Vollmer, Brittany Pettite, Marshall Thomas and Alexandra Denby for assistance in the field. Dave Post, who collected a large portion of the 2006 data, was a constant source of energy and entertainment during many long days. Tom Coe and the staff at the Warrensburg Hatchery provided critical support during transmitter implant surgeries and with daily fish care throughout the summers of 2005 and 2006. Howard Simonin and Alan Mack of the NYSDEC Rome and Caledonia hatcheries, provided access to fish and space for preliminary surgery studies and a trout condition study. Paul Bowser, Owen Baird, Travis Sanderson, Traci Barkley, and Jesse Lepak provided advice on surgical technique and blood sampling. Many residents of the Cedar River Road allowed access to the Cedar River over their private property. Special thanks to Elva Brown whose property we crossed to stock the trout and every tracking day after and Doug Wells who helped Dave and I out of a couple of jams. The Barton Mine Company allowed regular access to the Hudson River. The Gooley Club and Finch, Pruyn & Company, Inc. allowed regular access to the Indian River over their property during 2005 and 2006. The following local anglers provided information regarding current and historical fishing catch, sources of food and thermal refugia for trout in the study reaches and/or assistance catching fish: Patrick Sista, Larry Ames, Jim Nash, Gregg Hogancamp, and Wayne Failing. Wayne Failing and Pete Burns negotiated the iv Hudson River Gorge by raft so that we could track fish throughout this remote section of river. I received substantial assistance with study design and analysis from the Cornell Statistical Consulting Unit. Special thanks to Francoise Vermeylen for many hours of consultation. Dr. Evan Cooch and Dr. Patrick Sullivan also provided statistical advice. Justin Blanco helped with data manipulations and management. Rick Fenton and Rich Preall, who were my contacts at the NYSDEC provided support and critical feedback and helped steer my research to meet relevant needs of the state management agency. Walt Keller and Greg Nagle provided feedback on earlier drafts of this thesis. Fellow graduate students, especially members of the Kraft Lab, provided feedback on many aspects of this research and support throughout my time at Cornell. Dan Josephson provided guidance and logistical support, especially during field seasons. My family and friends have been a great source of encouragement and support. Finally I would like to thank my committee for their support and feedback: Barry Baldigo, Milo Richmond, Todd Walter and especially my advisor, Cliff Kraft. This research was supported by funds from the NewYork State Department of Environmental Conservation, the US Geological Survey – New York Water Science Center and New York Cooperative Fish & Wildlife Unit, and the Keickhefer Adirondack Fellowship. v TABLE OF CONTENTS Biographical Sketch Acknowledgements Table of Contents List of Figures List of Tables Chapter 1 – The effect of pulsed discharge events on thermal refugia use by brown trout in thermally marginal streams Introduction Site description Methods River temperature Trout behavioral thermoregulation Persistence Effects of recreational discharge events on thermal refuge habitat Effects of recreational discharge events on trout behavioral thermoregulation Empirical model selection Results River temperature Trout behavioral thermoregulation Persistence Effects of recreational discharge events on thermal refuge habitat iii iv vi viii x 1 6 8 8 8 12 13 14 16 17 17 19 20 20 vi Effects of recreational discharge events on trout behavioral thermoregulation Discussion River temperature Effects of recreational discharge events on thermal refuge habitat Effects of recreational discharge events on trout behavioral thermoregulation Implications Appendix A – Fate of stocked 2-year-old brown trout in the Upper Hudson River drainage Appendix B – Movement behavior and habitat use of stocked 2-year-old brown trout in the Upper Hudson River drainage Methods Results Discussion Appendix C – Qualitative analysis of the effect of pulsed discharge events on thermal behavior of 2-year-old brown trout in the Upper Hudson River drainage Appendix D – Estimation of missing river temperature data Appendix E – Table of statistics for individual trout implanted with radio transmitters References 28 43 44 45 46 52 55 61 63 64 72 75 79 84 88 vii LIST OF FIGURES Number 1.1 1.2 1.3 1.4a-b 1.5a-c 1.6 1.7 1.8 1.9 A.1 Description Page Map of the Upper Hudson River drainage including the study sites along the Cedar, Indian and Hudson Rivers. 7 Average daily ambient river temperature plotted over the length of the study for both 2005 and 2006 in the Cedar, Indian and Hudson Rivers. 17 Brown trout behavioral thermoregulation during the afternoon in the Cedar, Indian and Hudson Rivers during 2005 and 2006. 19 Persistence of brown trout in the Cedar, Indian and Hudson Rivers during the two study years. 21 Average, maximum and range of daily temperatures within three Hudson River tributary confluences plotted against mean daily discharge. 24 Comparison of the least-square mean estimates of the temperature difference between fish body and river between fish within and farther than 50 meters from a tributary under different ambient river temperatures for morning, midday and afternoon time periods in the Cedar River. 34 Comparison of the least-square mean estimates of the temperature difference between fish body and river between fish within and farther than 50 meters from a tributary for morning, midday and afternoon time periods in the Indian River. 38 Comparison of the least-square mean estimates of the temperature difference between fish body and river between release days and non-release days under different ambient river temperatures for the midday time period in the Indian River. 39 Comparison of the least-square mean estimates of the temperature difference between fish body and river between levels of the following variables: nearness to a tributary, river temperature and release day for the midday and afternoon time periods in the Hudson River. 42 Confirmed or inferred locations of transmitters at the conclusion of the study during 2005 and 2006. 57 viii A.2 B.1a-c B.2 B.3a-c C.1 C.2 D.1 D.2 D.3 The number of implanted fish alive and within the study reach plotted for each day of the study in 2005 and 2006 for Indian and Hudson Rivers and 2006 for the Cedar River. 60 Dispersal of brown trout in the Cedar, Indian and Hudson Rivers approximately one day, one week, and one and two months after stocking in 2006. 66 Activity level of brown trout in the Cedar, Indian and Hudson Rivers during the 2006 study period. 67 Bar charts of observed locations of brown trout in the Cedar, Indian, and Hudson Rivers plotted against distance within the study reach for 2006. 69 Categorization scheme for series of fish body temperature observations 1 ½ hours before and after the arrival of the release pulse at a fish’s location. 76 Summary of 108 individual series of fish body temperature observations 1½ hours before and after the arrival of the release pulse at a fish’s location within the Indian and Hudson Rivers. 77 Frequency distribution of the residuals of the interpolated values for missing temperature observations for the most downstream Hudson River temperature logger in 2005. 81 Frequency distribution of the residuals of the interpolated values for missing temperature observations for the most downstream Hudson River temperature logger in 2006. 82 Frequency distribution of the residuals of the interpolated values for missing temperature observations for the most downstream Cedar River temperature logger in 2006. 83 ix LIST OF TABLES Number 1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8 1.9 1.10 Description Page Sampling effort for trout telemetry study for 2005-06. 12 The maximum of the mean, maximum, and range of daily river temperatures averaged over consecutive days at a series of intervals following Wehrly et al. (2007). 19 Top ranked empirical models for the mean, maximum and range in daily temperature measured at three Hudson River tributary confluences. 22 ANOVA table, fit statistics and effects tests for the best supported multiple regression model for the mean daily temperature at three Hudson River tributary confluences. 23 ANOVA table, fit statistics, effects tests and least-square mean estimates for the best supported multiple regression model for the maximum daily temperature at three Hudson River tributary confluences. 26 ANOVA table, fit statistics, effects tests and least-square mean estimates for the best supported multiple regression model for the daily temperature range at three Hudson River tributary confluences. 27 Top ranked empirical models for the temperature difference between fish body and ambient river for each river / time-of-day combination. 29 Predictor weights for the three most important fixed effects for each river / time-of-day combination. 30 Covariance parameters for the random effects and test for significance of fixed effects for the top-ranked empirical models of the temperature difference between fish body and ambient river for all Cedar River / time-of-day combinations. 32 Least-square mean estimates of the temperature difference between fish body and ambient river for the top-ranked empirical model for each Cedar River / time-of-day combination comparing fish within and farther than 50 meters from a tributary at different values of river temperature. 33 x 1.11 Covariance parameters for the random effects and test for significance of fixed effects for the top-ranked empirical models of the temperature difference between fish body and ambient river for all Indian River / time-of-day combinations. 36 1.12 Least-square mean estimates of the temperature difference between fish body and ambient river for the top-ranked empirical model for each Indian River / time-of-day combination comparing fish within and farther than 50 meters from a tributary. 37 1.13 Least-square mean estimates of the temperature difference between fish body and ambient river for the top-ranked empirical model for midday in the Indian comparing release days and non- release days at different values of river temperature. 37 1.14 Covariance parameters for the random effects and test for significance of fixed effects for the top-ranked empirical models of the temperature difference between fish body and ambient river for all Hudson River / time-of-day combinations. 40 1.15 Least-square mean estimates of the temperature difference between fish body and ambient river for the top-ranked empirical model for midday and afternoon time periods in the Hudson River comparing fish within and farther than 50 meters from a tributary at different values of stream temperature and also comparing release days and non-release days during midday. 41 B.1 Mean distance from the stocking location for trout in the Cedar, Indian and Hudson River reaches at approximately one day, one week, one month and two months after stocking in 2006. 66 D.1 Dates of data missing from river temperature loggers. 80 D.2 ANOVA table, fit statistics, effects tests and parameter estimates with standard errors for the multiple regression model for the most downstream river temperature logger in the Hudson River in 2005. 80 D.3 ANOVA table, fit statistics, effects tests and parameter estimates with standard errors for the multiple regression model for the most downstream river temperature logger in the Hudson River in 2006. 81 D.4 ANOVA table, fit statistics, effects tests and parameter estimates with standard errors for the multiple regression model for the most downstream river temperature logger in the Cedar River in 2006. 82 E.1 Table of statistics for individual brown trout implanted with radio transmitters. 85 xi Introduction It is well established that fish move to areas of preferred temperatures to maximize growth, fitness, and survival (Ebersole et al. 2003a; Power et al. 1999; Torgersen et al. 1999; Garret and Bennett 1995). Yet salmonid species also inhabit streams with ambient summer temperatures approaching or exceeding lethal limits. Under these conditions, localized areas of cool water facilitate the persistence of coldwater fishes (Clapp et al. 1990; Ebersole et al. 2001; Berman and Quinn 1991; Matthews et al. 1994; Baird and Krueger 2001; Sutton et al. 2007). Pulsed discharge events impact salmonid behavior (Heggenes 1988b; Pert and Erman 1994; Bunt et al. 1999; Scruton et al. 2005) and alter available habitat in regulated rivers (Moog 1993; Valentin et al. 1996; Bain et al. 1998; Dare et al. 2002; Calles et al. 2007), including available thermal refugia (Sutton et al. 2007). No previous investigation has evaluated the impacts of pulsed discharge events on behavioral thermoregulation by salmonids. Critical temperature thresholds for brown trout (Salmo trutta) have been determined in laboratory studies. The low value of the “upper critical range”, the temperature range over which normal behavior of brown trout is disrupted, was estimated by Elliot (1994) to be 19°C. In a subsequent laboratory study of brown trout Elliot and Elliot (1995) identified the “upper incipient lethal temperature” – the maximum temperature that can be tolerated for one week – as 24.7oC and the “critical thermal maximum temperature” – the temperature that is lethal over a short period of time (tens of minutes) – as 29.9oC. Recent field-based estimates of brown trout thermal tolerances were identified using daily temperature means, maximums, and ranges over a series of exposure periods at locations where these fish were found within Wisconsin and Michigan streams (Wehrly et al. 2007). Daily temperature fluctuations enable salmonids to 1 survive in rivers with higher maximum temperatures than those without daily temperature fluctuations by providing intermittent periods of physiological stress and repair (Johnstone and Rahel 2003); however, the daily temperature range tolerated by brown and brook trout was shown by Wehrly et al. (2007) to decrease as mean river temperature increased into the critical range. Similarly, growth of fish is accelerated under conditions of low fluctuating temperatures, but depressed when temperatures fluctuate around values above a species’ thermal optimum (Jobling 1997). One of the benefits of field-derived characterizations of thermal tolerance is that they are based on the realized thermal niche of the species (Magnuson et al. 1979) and take into account sources of variation such as behavioral thermoregulation. Behavioral thermoregulation by stream-dwelling salmonids in localized cool water patches has been observed within the thermal mixing zones of tributary confluences (Kaeding 1996; Baird and Krueger 2003; Sutton et al. 2007), within stratified pools (Nielsen et al. 1994; Matthews et al. 1994; Elliot 2000; Baird and Krueger 2003; Tate et al. 2007), and locations associated with upwelling groundwater (Ebersole et al. 2001; Ebersole et al. 2003a). Temperature-sensitive radio transmitters have been used to determine the difference between an individual fish’s body temperature and the ambient river temperature. For example, Berman and Quinn (1991) reported that the body temperature of Chinook salmon (Oncorhynchus tshawytscha) in the northwestern United States was typically 2.5oC cooler than the ambient river temperature. Similarly, Baird and Krueger (2003) reported finding rainbow (Oncorhynchus mykiss) and brook trout (Salvelinus fontinalis) 2.3oC and 4oC cooler, respectively, than the Adirondack river they inhabited. Behavioral thermoregulation by fishes often varies diurnally and at different ambient river temperatures. In laboratory studies, brown trout presented with a range of temperatures selected the coolest water during daylight hours and warmer 2 temperatures at dawn and dusk (Reynolds and Casterlin 1979), which are time periods often associated with increased activity (Clapp et al. 1990; Bunnell et al. 1998) and feeding (Diana et al. 2004). Sutton et al. (2007) found an increase in the number of salmonids within thermal refuge areas throughout the day as river temperatures warmed, particularly when ambient river temperature exceeded 23°C. Similarly, Ebersole et al. (2001) found a peak in thermal refuge use by rainbow trout during the warmest part of the day. Baird and Krueger (2003) found that the temperature difference between the fish and the river for adult rainbow and brook trout was more negative when river temperatures were greater than 20°C. Similarly, Matthews et al. (1994) found that rainbow trout did not seek cold water refuge at temperatures below 19.3°C, and brown trout entered relatively cool tributaries when reservoir temperatures reached 19-20°C (Garret and Bennett 1995). A number of studies have found that salmonids also select temperatures that exceed their reported optimal range and have attributed this to the importance of other physical or chemical habitat variables, competition, feeding, predator avoidance or movements to more suitable habitat (Jobling 1981; Matthews et al. 1994; Elliot 2000; Ebersole et al. 2001; Baird and Krueger 2003; Sutton et al. 2007). Additionally, human disturbances such as dams or flow regulation may reduce the quantity of suitable thermal habitat (Poole and Berman 2001). Sutton et al. (2007) described one such mechanism where localized cool water patches in a stream were constricted or diluted by water discharged from upstream reservoirs. Several studies have investigated impacts of hydro-peaking on salmonid movement, habitat selection, feeding, growth, and stranding (Heggenes 1988b; Bradford 1997; Bunt et al. 1999; Saltveit et al. 2001; Flodmark et al. 2002; Scruton et al. 2003; Flodmark et al. 2004; Scruton et al. 2005; Heggenes et al. 2007). In studies of yearling and two-year-old brown trout, individuals moved closer to the river 3 margins during peak releases and into areas with woody debris and velocities slower than those selected during natural flows (Bunt et al. 1999). When little cover is available, this behavior may expose individuals to predation. Heggenes et al. (2007) found no consistent impact of peaked flows on the areas used by and the movements of brown trout, yet observed a non-significant trend for increasing home ranges and movements in relation to higher artificial flows. Increased activity due to peaking events would increase energetic demand of coldwater fishes in thermally marginal streams. In a study conducted in streams with temperatures within the optimal range for brown trout, Flodmark et al. (2004) concluded that low and fluctuating flows and low flows combined with fluctuating temperature may be detrimental to the growth rate of juvenile brown trout. Scruton et al. (2005) and Flodmark et al. (2004) noted an increased energetic cost for fish that change position in response to these flow and temperature stressors. Flodmark et al. (2002) found that juvenile brown trout, subject to rapid daily flow reductions, initially showed an acute stress response in blood cortisol levels, but after four days the response was no longer present. If the latter resulted from compensation (avoidance of the stressor), these results suggest that individuals would experience greater energetic costs that may produce long-term negative effects while searching for suitable habitat (Flodmark et al. 2002). The coupling of summer low flows and high temperatures with increased fluctuations in discharge, and possibly temperature, would likely create a very stressful environment for coldwater fish. In this study we examined the variation in brown trout behavioral thermoregulation within three streams of the Hudson River drainage, the Cedar, Indian and Hudson Rivers, the latter two of which are impacted by recreational discharge events from an upstream dam. Extensive whitewater reaches in the Indian and Hudson 4 Rivers provide a setting for a commercial rafting industry operating from April through October. To enhance rapids and enable summer rafting, regular releases are made from a shallow impounded lake on the Indian River, 4.5 km upstream from its confluence with the Hudson. The key questions driving this study were whether pulsed discharges decrease the quality and quantity of coldwater fish habitat and alter the thermal behavior of brown trout – an important sport fish and actively managed species in this river. We were particularly interested in evaluating the presence and potential impact of thermal refugia dilution (Sutton et al. 2007). In order to evaluate the influence of dam releases on brown trout thermal behavior, we first characterized thermal conditions in the main river channels of the three study reaches and in low-order tributary confluences. We then evaluated the degree of brown trout behavioral thermoregulation in each reach by investigating changes in the differences between trout body temperature and ambient river temperature in affected and reference reaches and under release and non-release conditions. In addition, the possible effects of recreational releases on trout survival were assessed using inference from indirect data. We hypothesized the following: 1) Brown trout would seek water that would enable them to maintain body temperatures below those found to be physiologically stressful. 2) Brown trout body temperature would deviate most from ambient river temperature during the warmest part of the day. 3) Recreational releases would alter the characteristics of the available thermal refugia. 4) Recreational releases would alter the thermal behavior of brown trout and reduce differences between fish body temperature and ambient river temperature. 5 Site Description The Upper Hudson River drainage is located in the southeastern Adirondack Mountains of New York in Hamilton, Essex and Warren counties. The surrounding land cover is mature second-growth northern hardwood and mixed northern hardwood-conifer forests. The study was conducted in three reaches (Figure 1.1), one in the mainstem Hudson River (5th order) and two in large tributaries, the Indian River and the Cedar River (3rd order). These reaches support small transitory populations of native brook trout and seasonal populations (hatchery origin) of brown trout and rainbow trout. The gradient within all reaches was moderate, ranging from 0.006 in the Hudson River reach to 0.014 in the Indian River reach, with boulder-cobble dominated (Indian and Hudson reaches) or cobble-gravel dominated (Cedar reach) substrate. Both the Cedar and Indian River reaches were downstream of impoundments (Figure 1.1). The hydrological regime of the Indian River reach is manipulated from April to October by short duration discharge events produced for recreational boating. These peaking events occur daily in the spring and four days each week in the summer. Daily discharge measured immediately downstream of the dam averaged 7.3 and 12.5 cubic meters per second (cms) during June – September in 2005 and 2006, respectively (Baldigo et al. in prep). Recreational discharge events increased flows over a 30–minute period to an average of 39.3 cms and persisted for approximately 1 ½ hours before declining, on average, 1.4 cms below the flow level at the start of discharge. This subsequent drop was associated with the recharge of Lake Abanakee, which was also influenced by an upstream regulated dam (Indian Lake Dam) (Baldigo et al. in prep). Recreational discharge events also increased stage and discharge within the Hudson River reach located 20-30 km downstream from the dam, beginning at the 6 Boreas River and ending in the town of North Creek (Figure 1.1). No biologically significant differences between water temperature on release days and non-release days were found at any of the measured sites (Baldigo et al. in prep). The upstream end of the Cedar River reach began at the Wakely Dam, an unregulated dam, and ended where Route 28 crosses the river. A second dam, the Cedar River Dam (an unmaintained mill dam) is located in the middle of the reach. Cedar River Cedar River Dam New York Hudson River 500 m Boreas River Cedar River Wakely Dam Indian River Hudson River 2 A B N 4 km Indian River Lake Abanakee Dam 500 m 1 Dam 1 km C Stocking Location River temperature Logger Thermal refugia logger (A, B, C) Stationary radio-telemetry receiver location Figure 1.1. Map of the study area. Length of each river reach that was tracked is indicated with a bold line. Inset maps of each river reach display the section of reach where 95% of observations were obtained. Recreational flow releases originate at the Lake Abanakee Dam on the Indian River. Hudson River thermal refuge areas monitored were at the confluences of Griffin Brook (A), Raquette Brook (B), and Balm of Gilead (C). 7 Methods River temperature Water temperature and stage were recorded at fifteen minute intervals with data logging pressure transducers in the Indian (three loggers) and Hudson River (two loggers) during both years throughout the entire telemetry study (Figure 1.1). In 2006, ambient river temperatures in the Cedar River study reach were recorded at fifteen minute intervals at three locations with Stowaway loggers (Figure 1.1). During the study period, the maximum, minimum and average temperature within a reach during each fifteen minute time interval was determined by averaging the measurements taken by all loggers within that reach. The maximum of the moving average of mean, maximum, and range of daily temperatures for consecutive days were calculated at a series of intervals (1 day, 7 days, 21 days, and 63 days) following Wehrly et al. (2007). These additional calculations allowed us to make comparisons regarding both the magnitude and duration of thermal stress. Trout behavioral thermoregulation Temperature sensitive transmitters were used to monitor the location and body temperature of 30 hatchery reared, two-year-old brown trout during summer 2005 and 47 similar fish during summer 2006 (mean total length ± SE = 378.5 ± 3.4mm in 2005 and 371.7 ± 1.7 mm in 2006). Trout were stocked at three sites; two affected by recreational flow releases (within the Indian and Hudson Rivers) during both years and an additional site unaffected by dam releases (within the Cedar River) during 2006 only. Stocking occurred on July 25 in 2005 and June 14 in 2006 (locations indicated in Figure 1.1), and trout were monitored up to six days each week until August 18 during both study years. Two large flooding events occurred during 2006 and data acquired 8 during these events, defined as daily discharge greater than or equal to 73.6 cms at the USGS gage station on the Hudson River at North Creek (USGS 01315500), were excluded from these analyses. Model F1815 (battery life = 42 days, 9 grams) and model F1820T (battery life = 140 days, 10grams) temperature-sensitive radio transmitters (Advanced Telemetry Systems, Isanti, MN) were implanted in fish in 2005 and 2006, respectively (manufacturer specified accuracy of +/- 0.25°C and precision of +/- 0.5°C). Laboratory tests conducted on transmitters used in 2006 found that the average (±SE) difference between the temperature measured by a transmitters and a calibrated temperature meter was 0.01 +/- 0.02°C. The mean difference for any individual transmitter exceeded 0.2°C in only one case in which the difference was 0.34°C, therefore that value was added to all observations from this fish. Transmitters were also evaluated by transferring them from cool to warm water and measuring the time until the temperature stabilized. On average, transmitters warmed to an accuracy of 0.04 +/- 0.01°C in 154 +/- 3 seconds (N=21). Although no laboratory tests were conducted to evaluate the rate of fish body temperature increase relative to ambient water temperature change, or the corresponding accuracy of measuring these changes, field observations showed transmitters recording fish body temperatures increasing at a rate of 0.1°C per minute. Transmitters were implanted by anesthetizing each fish, inserting a transmitter into the abdominal cavity and sealing the incision with sutures using methods similar to the shielded-needle technique (Ross 1982; Summerfelt and Smith 1990; Wooster and Bowser 1993). The mean ratio of transmitter to fish weight was 1.3% in 2005 and 1.5% in 2006. Fish were held at the hatchery for recovery for two weeks before being released into the rivers. Additionally, dummy transmitters were implanted into fish (N = 5 in 2005 and N = 10 in 2006) and held at the hatchery to assess potential mortality 9 or unusual behavior caused by the surgeries and to evaluate potential transmitter expulsion. All fish held in the hatchery survived past the conclusion of the telemetry surveys and exhibited normal behavior, and none expelled transmitters until after the field tracking efforts were completed. Two additional study trout were stocked into the Hudson River on July 10, 2006 to replenish the population of study fish that were dying or disappearing more rapidly than in either the Cedar or Indian Rivers. During each day of sampling we attempted to locate and obtain multiple body temperature readings from each fish at all sites. The entire reach of each river was searched whenever possible, though weather conditions or logistical constraints infrequently prevented complete surveys. Data were collected by two methods. The primary means was manual collection by walking or driving the banks of the study reaches with a 3-element Yagi antenna and an ATS RS4500 data-logging receiver, previously set to aerial scan mode, and collecting one data point every second. Data from implanted transmitters were also collected by a fixed location ATS RS4500 datalogging receiver (2006 only) installed on the Hudson River. The scan time and record interval were set such that a temperature would be recorded every five minutes if fish were in range (roughly 1 km) throughout an entire 24-hour daily period. For the first ten days after stocking, the fixed receiver was positioned approximately 10 km downstream from the stocking location (Figure 1.1; 1∗) to identify any fish that exited the study reach. After no such movements were observed during this time period, the fixed receiver was moved (June 24, 2006) to a location approximately 1 km upstream from the stocking location (Figure 1.1; 2∗) and within range of one major and two minor tributaries. The receiver was terminated on July 18, 2006. To characterize changes in trout thermal behavior throughout the day, observations during the morning (5:00-8:59 EST), midday (9:00-12:59 EST) and afternoon (13:00-16:59 EST) time periods were collected in the Cedar River (no recreational flow release) study reach. 10 For the reaches affected by flow releases we attempted to collect daily body temperatures for each fish both before and during a release (or during these same time periods on days when no release occurred), and after a release in the Indian River when daylight permitted. The release generally passed through each reach within one of the designated periods for both the Indian River (midday) and Hudson River (afternoon). From the collected data, we designated the median body temperature recorded for each fish during each time period as a sampling observation. These observations were paired with a measured river temperature at the nearest logger (median, minimum and maximum distances of fish from nearest logger were 401, 5 and 9760 meters, respectively) at the 15-minute interval closest in time to when the fish body temperature was recorded. For study trout in the Indian and Hudson Rivers, body temperatures were assigned a condition of either release or non-release based on whether the stage measurement at that 15-minute time interval indicated the presence of a flow release pulse. On release days, observations made when the fish was not experiencing release conditions during the release time period were eliminated (midday in the Indian River and afternoon in the Hudson River). The Hudson River dataset included observations recorded using the fixed receiver when manual tracking data were not available (Table 1.1). We defined periods of behavioral thermoregulation based on observations of brown trout body temperature at least 1°C cooler than the ambient river temperature. Because brown trout have been found to more frequently use thermal refugia during the warmest part of the day (Ebersole et al. 2001; Sutton et al. 2007), we used only observations during the afternoon when the ambient river temperature was ≥ 20°C. Additionally, we excluded observations taken on release days because a portion of Hudson River observations would have been influenced by the recreational discharge 11 events. We report behavioral thermoregulation by the study trout as the ratio of observations in which trout were behaviorally thermoregulating to the number of total observations. Table 1.1. Sampling effort for trout telemetry study summarized for 2005-06. Note that total number of observations includes up to three observations per fish per day. Year 2005 2005 2006 2006 2006 Study reach Total number of days tracked Number of flow release days tracked Indian River Hudson River Indian River Hudson River Cedar River 17 15 45 50 46 8 9 21 24 21 Total number of telemetry observations 183 103 659 380 555 Number of telemetry observations during flow release days 83 70 393 233 329 Persistence The end date for each fish with an implanted transmitter was defined as the first day when that individual was no longer observed alive within the study reach. In 2006 the transmitters were equipped with mortality sensors that produced a different signal when a transmitter remained still for more than eight hours, and the end date was defined as the day of the first observation prior to receiving a mortality signal. If the location of that transmitter had not changed for multiple days prior to the mortality signal, the end date was determined as the first day at the final observed location. Similarly, if no mortality signal was emitted and there was no change in movement from the final observed location, the first day at that location was determined to be the end date. If the signal indicated, either by temperature or location, that the transmitter 12 was out-of-water, the end date was identified as the day that this condition was first observed. Transmitters found lying within the main river channel were assumed to be derived from a dead fish. Persistence was calculated as the number of days that each fish was alive and remained within the study reach (Bettinger and Bettoli 2002). Estimates of median persistence were determined using the SAS LIFETEST procedure, which accounts for observations that were “censored” (i.e. trout that survived past the conclusion of the study). Comparisons were made between rivers within each year, and a Wilcoxon statistic was used to test the homogeneity between survival curves (Allison 1995). In this analysis lost transmitters may be considered to result from mortality and (or) emigration, i.e., the fish was removed from the system by some means. Effects of recreational discharge events on thermal refuge habitat The temperature regime of three tributary confluences to the Hudson River (Griffin Brook, Raquette Brook, and Balm of Gilead; Figure 1.1) were recorded in 2006 using a single Stowaway temperature logger per tributary that was secured near the substrate at a location where study trout had been observed in 2005. These loggers measured water temperature within the mixing zone where cool tributary water entered the warmer mainstem river. The effects of the recreational discharge events on the daily temperature mean, maximum, and range were assessed using multiple linear regressions. We developed a set of 33 models that included environmental variables likely to influence the thermal regime within these refuge areas. The predictor variables included maximum daily mainstem river temperature, mean daily discharge (measured at USGS Gage 01315500 in North River, NY), release day (release or nonrelease), tributary (to determine differences between the three locations), time (to account for autocorrelation between adjacent measurements), and the interactions 13 between combinations of release day, mean daily discharge, and tributary. An interaction between variables would indicate that the correlation of the dependent variable with one of the interacted independent variables varied based on the value of the other variable in the interaction. To further examine significant effects of release day, multiple comparisons with a Bonferroni correction were performed (Zar 1996). When included in an interaction term with categorical variables, least-square mean estimates were calculated holding the continuous variables constant at a value near the upper and lower third of the data range in order to discern the trends at high and low values of the variable. Effects of recreational discharge events on trout behavioral thermoregulation The relationship between recreational discharge events and the temperature difference (TD) between trout body and ambient river, where a negative TD indicated trout body temperatures cooler than the river, was assessed using multilevel models (Snijders and Bosker 1999). For each river / time-of-day combination, we developed a set of multilevel models incorporating the effect of release day and other factors we hypothesized would impact trout behavioral thermoregulation (29, 103 and 51 models each within Indian, Hudson and Cedar River reaches, respectively). Analyses were run separately for each river / time-of-day combination in order to simplify models and increase ease of interpretation. In each of these models the response variable was the TD. Multilevel models are effective analysis tools for longitudinal data (Singer 1998). To account for repeated measurements from individual fish and multiple measurements taken on a single day, fishID (unique individual fish identifier) and day were entered as random effects in every model (Littel et al. 1996; Snijders and Bosker 1999). We included combinations of the following fixed effects in each model: day-today temperature variation, using river temperature at the time of observation as the 14 metric (rivT); mean daily discharge, measured at USGS Gage 01315500 in North River, NY (mdd); nearness to a tributary, where a location within fifty meters was categorized as “near” and all other locations were not (ntrib); release day using two categories, release and non-release (rel); and, where sufficient data existed, the interaction between combinations of these variables (indicated with *). Interactions between nearness to a tributary and other variables were not possible for the Indian River because of insufficient numbers of observations of trout near a tributary. An additional variable was added to the Hudson River models. The distance of the fish from the nearest river temperature logger (dlog) was included to account for potential bias due to habitat being more variable within the study reach and the distance between the river temperature loggers being greater in the Hudson River than the other two reaches. Analysis of multilevel models was performed using the MIXED procedure (Littell et al. 1996) in SAS (SAS Institute Inc., Cary, NC). Multilevel models were run on all reported models for each river / time-of-day combination to obtain parameter estimates and significance of fixed effects. Unconditional covariance parameters for the random effects were estimated with a multilevel model with no fixed effects. These values can be used to interpret the amount of explainable variation accounted for by adding fixed effects to the model, where a reduction in a covariance parameter within a conditional model (one that includes fixed effects) indicates that some of the explainable variation at that level was described with the added fixed effects (Snijders and Bosker 1999). To further examine significant effects, differences between leastsquare mean estimates using a Bonferroni correction for multiple comparisons were calculated (Zar 1996). When included in an interaction term with categorical variables, estimates of the least-square mean were calculated holding the continuous variables 15 constant at a value near the upper and lower third of the data range in order to discern the trends at high and low values of the factor. Empirical model selection To compare the relative support given by the data for each of the models in a given model set, we used Akaike’s Information Criterion (AIC; Burnham and Anderson 2004) model selection techniques. Models within each set were ranked by corrected AIC value (AICc), where the best model, or the model with the most support from the data, had the lowest AICc value. To make initial comparisons between models, the ∆ AICc was calculated by subtracting the AICc value of the best model from each of the other models in the set. AICc weight (wi), a normalized likelihood, was calculated to provide a stronger measure of relative support for each model (Burnham and Anderson 2004). The AICc weight was interpreted as the probability that a given model was the best model within the model set. For the thermal refugia model sets and for each river / time-of-day combination of the TD model sets, we focus our discussion on only the best supported model, but report the best three models or all models with considerable support (∆ AICc < 7) (Burnham and Anderson 2004), whichever is more inclusive. The relative importance of fixed effects within a TD model set was also determined by summing the AICc weights for each model that contained a given fixed effect to determine its predictor weight (w+(j)) (Burnham and Anderson 2002). 16 Results River temperature Average daily temperatures for all study reaches peaked from late July to early August (Figure 1.2). During the 2005 study period, the mean water temperature exceeded 20°C on all survey dates, exceeded 25°C on approximately 50% of survey dates and seldom dropped below 20°C in the Indian and Hudson Rivers. In 2006, the river temperature exceeded 20°C during approximately 90% of the survey dates in both the Indian and Hudson Rivers and 80% of the time in the Cedar River. Water temperature exceeded 25°C in all three river reaches approximately 10% of the days. Temperature (°C) 30 28 26 24 22 20 18 16 14 12 10 9JUN 19JUN 29JUN 9JUL 19JUL 29JUL 8AUG 18AUG 28AUG Hudson River 2005 Indian River 2005 Hudson River 2006 Indian River 2006 Cedar River 2006 Figure 1.2. Average daily ambient river temperature plotted over the length of the study in both 2005 and 2006 for the Hudson, Indian and Cedar Rivers. In 2005 both the Hudson and the Indian Rivers exceeded the upper tolerance maximum and mean temperatures for brown trout identified by Wehrly et al. (2007) for both the 7-day and 21-day exposure categories by up to 3.3ºC. The Hudson River also exceeded the 1-day thresholds (Table 1.2). The mean thresholds were exceeded 17 by similar magnitudes for both study reaches, but the maximum thresholds were exceeded by greater magnitudes in the Hudson River. The daily temperature range in the Hudson River reach was approximately twice that of the Indian River reach for all exposure categories in 2005. In 2006, the mean and maximum thresholds for shorter exposure periods (1 and 7-day), identified by Wehrly et al. (2007), were not exceeded in any of the reaches, with the exception of the 7-day average in the Indian River (exceeded tolerance by 0.4°C) (Table 1.2). However, tolerance thresholds for the longer exposure periods were surpassed. All rivers exceeded both mean and maximum thresholds for 21-day exposure, but by no more than 0.3ºC for the maximum daily threshold and 1.4ºC for the average daily threshold. Both the Indian and Hudson Rivers exceeded the daily mean 63-day exposure threshold by 0.9ºC and 0.6ºC, respectively. The maximum daily temperatures for all exposure categories for all study reaches were similar. The highest mean daily value was found in the Indian River and the lowest mean daily value was found in the Cedar River for all exposure durations except the 1day category. Similar to 2005, the 2006 daily temperature ranges were greater in the Hudson River reach than the Indian River reach, but by a lesser degree. The Cedar River temperature range generally surpassed both of the other rivers for all exposures lengths during 2006. 18 Table1.2. The maximum of the mean, maximum, and range of daily river temperatures averaged over consecutive days at a series of intervals (1 day, 7 days, 21 days, and 63 days) following Wehrly et al. (2007). 2005 2006 1day 7day 21day 1day 7day 21day 63 day Cedar River Max Ave Range not measured 26.39 24.34 24.51 22.59 24.90 22.61 22.77 20.95 5.72 4.31 3.47 3.27 Indian River Max 26.72 26.17 25.10 25.63 24.51 24.22 22.78 Ave 24.96 24.64 23.70 24.8 23.71 23.51 21.89 Range 3.30 2.90 2.54 4.41 3.42 2.20 1.66 Hudson River Max 29.11 27.93 26.27 26.16 24.37 24.46 22.93 Ave 25.52 24.84 23.82 24.82 23.19 23.21 21.63 Range 8.81 6.42 4.98 5.78 4.03 2.85 2.47 Wehrly et al. 2007 Max Ave see below 27.6 25.3 25.4 23.3 24.2 22.1 22.9 21.0 Trout behavioral thermoregulation During the afternoon on non-release days (i.e. the warmest time of day) trout body temperatures were at least 1°C cooler than the river in 38% of the observations from the Cedar River, 29% from the Hudson River, and 4% from the Indian River when the ambient river was warmer than 20°C (Figure 1.3). Proportion of observations showing thermoregulation 1 0.8 0.6 0.4 N=99 0.2 0 Cedar N=66 Hudson N=144 Indian Figure 1.3. Brown trout behavioral thermoregulation during the afternoon, reported here as the number of observations of trout with body temperature at least 1°C cooler than the river divided by the total observations. Only observations on non-release days when the ambient river temperature was greater than 20°C are included. 19 Persistence Study trout in the Hudson River persisted for the shortest time (median duration: 2005 = 12 d, 2006 = 23 d). Trout in the Indian River persisted for a slightly longer duration (median duration: 2005 = 16 d, 2006 = 36 d) and trout in the Cedar River persisted for the longest duration (2006 median = 67 d) (Figure 1.4a-b). Persistence within the Hudson and Indian River study reaches were not significantly different in 2005 (x2 = 0.23, p = 0.06) or in 2006 (x2 = 0.97, p = 0.33). In 2006, persistence in both the Indian (x2 = 4.12, p = 0.04) and Hudson Rivers (x2 = 8.11, p < 0.01) was significantly different from the reference Cedar River. At the end of the 2005 season, only one study fish remained in the Hudson River and no study fish remained in the Indian River. At the end of the 2006 season, one study fish remained in the Indian River, two remained in the Hudson River and eight remained in the Cedar River (Appendix E). Effects of recreational discharge events on thermal refuge habitat A significant difference was found in the daily temperature maximum and range within the three monitored Hudson River tributary confluences (where cool tributary water mixes with the warmer mainstem river) between release days and nonrelease days. Release day did not have an effect on the daily mean temperature, but on release days the maximum daily temperature and the daily temperature range were greater than on non-release days. The magnitude of this effect was smallest within the Raquette Brook confluence and greater at low base flows within all monitored confluences. 20 Persistence (days) 80 70 60 50 40 30 20 10 0 Indian a. 2005 Hudson 80 70 60 50 40 30 20 10 0 Cedar b. 2006 Indian Hudson Figure 1.4a-b. Persistence (number of days a trout was alive and within the study reach) of brown trout in the Cedar, Indian and Hudson Rivers during the two study years. The Cedar River was only studied in 2006. Crosses represent the median, vertical lines extend to the maximum and minimum values, and grey boxes represent the middle 50% of the observations. In evaluating the mean daily temperature within the monitored tributary confluences the model with the best support from the data (Table 1.3) included the following parameters: the interaction between tributary and mean daily discharge (F2 = 11.18, p < 0.01), the interaction between tributary and release day (F2 = 0.26, p = 0.77), and the maximum daily temperature within the mainstem river (F1 = 197.71, p < 0.01) (Table 1.4). The best model explained approximately 80% (R2 = 0.83, p< 0.01) of the variation in mean daily temperature, with 1 ½ times more support than the second best model, but 30 times more support than the third (Table 1.3). The second 21 best model omitted the interaction between tributary and release day, which was not significant and provided little additional explanatory power in the best model. Release day was not a significant factor explaining variation in the mean daily temperature at the tributary confluences, but mean daily temperature at the confluences was positively correlated with maximum daily mainstem river temperature (Table 1.4 and Figure 1.5a). The mean daily temperature at each confluence was significantly different than that found at the other two confluences at both low (20 cms) and high (50 cms) values of mean daily discharge. Table 1.3. Top ranked empirical models for the mean, maximum and range in daily temperature measured at three Hudson River tributary confluences and determined using AIC model selection techniques. The top three or all models with a ∆AIC < 7, whichever is more inclusive, model likelihood (£), and R2 are reported with value. Maximum the AICc, ∆AICc, AICc weight (wi), daily mainstem river temperature = mxT, tributary = trib, release day = rel, mean daily discharge = mdd, and * indicates an interaction. Mean Maximum Range Model AICc ∆ AICc wi £ R2 trib*rel, trib*mdd, mxT 533.1 0.0 0.58 1.00 0.83 trib*mdd, mxT 533.9 0.8 0.39 0.67 0.83 trib*rel, mdd*rel, trib*mdd , mxT 539.9 6.8 0.02 0.03 0.83 trib*rel, mdd*rel, mxT trib*rel, mdd*rel, trib*mdd, mxT mdd*rel*trib, mxT 572.3 575.9 580.9 0.0 0.85 1.00 0.70 3.6 0.14 0.17 0.72 8.6 0.01 0.01 0.73 trib*rel, mdd*rel trib*rel, mdd*rel, trib*mdd trib*rel, mdd*rel, mxT trib*rel, mdd*rel, time trib*rel, mdd*rel, trib*mdd, mxT mdd*rel*trib trib*rel, mdd*rel, trib*mdd, time 561.7 564.3 564.5 566.6 567.2 568.3 568.6 0.0 0.57 1.00 0.63 2.6 0.16 0.27 0.66 2.8 0.14 0.25 0.64 4.9 0.05 0.09 0.64 5.5 0.04 0.06 0.66 6.6 0.02 0.04 0.67 6.9 0.02 0.03 0.67 22 Table 1.4. ANOVA table, fit statistics and effects tests for the AIC-selected, best supported multiple regression model for mean daily temperature at three Hudson River tributary confluences. Source DF Sum of Squares Mean Square F Value Pr > F Model 9 1115.25 123.92 78.87 <0.01 Error 145 227.83 1.57 Corrected Total 154 1343.07 R2 CV Root MSE Mean of average (°C) 0.83 6.81 1.25 18.41 Source trib mdd rel mxT trib*rel mdd*trib DF Type III SS 2 118.49 1 452.77 1 1.29 1 310.64 2 0.82 2 35.15 Mean Square 59.24 452.77 1.29 310.64 0.41 17.57 F Value 37.71 288.17 0.82 197.71 0.260 11.18 Pr > F <0.01 <0.01 0.37 <0.01 0.77 <0.01 23 daily temperature range ( °C) maximum daily temperature (°C) average daily temperature (°C) 28 a. 26 24 22 20 18 16 14 12 0 b. 28 26 24 22 20 18 16 14 12 0 12 c. 10 8 6 4 2 0 0 20 40 60 mean daily discharge (cms) 80 20 40 60 mean daily discharge (cms) 80 20 40 60 mean daily discharge (cms) 80 Griffin Brook Raquette Brook Balm of Gilead Non-release days Release days Non-release best fit Release best fit Figure 1.5a-c. Average, maximum and range of daily temperatures at three Hudson River tributary confluences plotted against mean daily discharge. Release day values are indicated with grey symbols and non-releases are indicated with black symbols. Griffin Brook (A in Figure 1.1) is represented with squares; Raquette Brook (B in Figure 1.1) is represented with triangles; Balm of Gilead (C in Figure 1.1) is represented with circles. 24 The model of maximum daily temperature within tributary confluences with the most support from the data (Table 1.3) included the following parameters: the interaction between tributary and release day (F2 = 8.73, p < 0.01), the interaction between mean daily discharge and release day (F1 = 33.70, p < 0.01), and the maximum daily temperature in the mainstem (F1 = 117.85, p < 0.01) (Table 1.5). This model, which explained 70% (R2 = 0.70, p < 0.01) of the variation in maximum daily temperature within the tributary confluences, had more than six times more support than the second best model (Table 1.3). Maximum daily temperature at the confluences was positively correlated with maximum daily mainstem river temperature. The least-square mean (LSMean) values for both Griffin Brook and Balm of Gilead differed significantly on release and non-release days, where the least-square mean estimates for maximum temperature were higher on release days (Table 1.5 and Figure 1.5b). Although the trend of increased temperature on release days was evident in the Raquette Brook confluence, it was more moderate and not significant. Maximum daily temperatures were also greater on release days than non-release days under both low and high base flow conditions. At high base flow the differences were less extreme. The model with the most support from the data for the range in daily temperature within the tributary confluences (Table 1.3) included: the interaction between tributary and release day (F2 = 10.52, p < 0.01) and the interaction between mean daily discharge and release day (F1 = 35.57, p < 0.01) (Table 1.6). This model explained 63% of the variation in daily temperature range (R2 = 0.63, p < 0.01) and had almost four times more support than the next best model (Table 1.3). The leastsquare mean estimates of temperature range for all three tributary confluences were significantly greater on release days than non-release days (Table 1.6 and Figure 1.5c). Similar to results for the maximum daily temperature, the difference was smallest 25 within the Raquette Brook confluence. The range was very similar on non-release days at both high (50 cms) and low (20 cms) values of mean daily discharge. On release days the least-square mean estimates were significantly greater for both levels of discharge than on non-release days, but the increase was more substantial at low flows. Table 1.5. ANOVA table, fit statistics, effects tests and least-square mean estimates for the AIC-selected, best supported multiple regression model of maximum daily temperature at three Hudson River tributary confluences. Source DF Sum of Squares Mean Square F Value Pr > F Model 8 714.97 89.37 41.91 <0.01 Error 146 311.32 2.13 Corrected Total 154 1026.30 R2 CV Root MSE Mean of maximum (°C) 0.70 6.99 1.46 20.88 Source trib mdd rel mxT mdd*rel trib*rel DF Type III SS 2 97.61 1 130.81 1 165.01 1 251.29 1 71.85 2 37.22 Mean Square 48.81 130.81 165.01 251.29 71.85 18.61 F Value 22.89 61.34 77.38 117.85 33.70 8.73 Pr > F <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 tributary Griffin Brook Raquette Brook Balm of Gilead non-release LSMean (°C) 18.16 21.39 19.54 release LSMean (°C) 21.39 22.15 21.67 Pr > F <0.01 0.21 <0.01 mean daily discharge non-release LSMean (°C) relsease LSMean (°C) Pr > |t| low (20 cms) 17.91 21.46 <0.01 high (50 cms) 20.92 21.93 <0.01 26 Table 1.6. ANOVA table, fit statistics, effects tests and least-square mean estimates for the AIC-selected, best supported multiple regression model of daily temperature range at three Hudson River tributary confluences. . Source DF Sum of Squares Mean Square F Value Pr > F Model 7 512.12 73.16 36.22 <0.01 Error 147 296.89 2.02 Corrected Total 154 809.01 R2 CV 0.63 37.66 Root MSE 1.42 Mean of range (°C) 3.77 Source trib mdd rel mdd*rel trib*rel DF Type III SS 2 8.22 1 139.17 1 207.22 1 71.84 2 42.49 Mean Square 4.11 139.17 207.22 71.84 21.25 F Value 2.04 68.91 102.6 35.57 10.52 Pr > F 0.13 <0.01 <0.01 <0.01 <0.01 tributary Griffin Brook Raquette Brook Balm of Gilead non-release LSMean (°C) 1.91 2.64 2.18 release LSMean (°C) 5.74 3.87 5.10 Pr > F <0.01 <0.01 <0.01 mean daily discharge low (20 cms) high (50 cms) non-release LSMean (°C) 2.54 2.04 relsease LSMean (°C) Pr > |t| 6.71 <0.01 3.66 <0.01 27 Effects of recreational discharge events on trout behavioral thermoregulation In the reference Cedar River reach, which was not subject to pulsed discharge events, the most important variables explaining variation in the difference between the body temperature of radio transmitter implanted fish and ambient river temperature during all time periods was the interaction between ambient river temperature and whether a trout was near a tributary (Table 1.7). In most cases the temperature difference was more negative when trout were near tributaries and when ambient river temperature was within the upper critical range for brown trout (> 19°C). The same was true for the Hudson River reach on days and time periods unaffected by releases. Similarly, in the Indian River reach, nearness to a tributary was the most important parameter on days and time periods unaffected by releases. Although release day was included in well supported models in these unaffected river / time-of-day combinations, the relative importance of this factor was low (Table 1.8). When a reach was inundated by release flows, release day or an interaction term including release day was at least as important as any other variable in explaining temperature differences. In both the Cedar and Hudson River reaches, mean daily discharge was also important during the afternoon time period. In the reference Cedar River, the best supported model accounting for behavioral thermoregulation in brown trout during the morning time period included a single fixed effect, the interaction between river temperature and nearness to a tributary (Table 1.7). This model had almost five times more support than the next best model, and the fixed effect explained a significant amount of the variability (F1,170 = 48.76, p < 0.01) (Table 1.9). Furthermore, a significant difference was found between the least-square mean estimates for the temperature difference between the fish and the river (TD) for observations within and farther than 50 meters from tributaries when the river temperature was held constant within the model at both low 28 Table 1.7. Top ranked empirical models for the temperature difference between fish body and ambient river for each river / time-of-day combination. The top three models or all models with a ∆AIC < 7, whichever is more inclusive, are reported with AICc, ∆AICc, AICc weight (wi) and the model likelihood (£). River temperature = rivT, nearness to a tributary = ntrib, release day = rel, mean daily discharge = mdd, distance from nearest logger = nlog and * indicates an interaction. Main effects and two-way interaction terms that build the reported models were included in model calculations, but are not shown in the table. “Release” indicates the time period during which the release pulse travels through a given river reach. Cedar River Morning Midday Afternoon Rank 1 2 3 Model rivT * ntrib rivT * ntrib, rel rivT * rel * ntrib 1 rivT * ntrib 2 rivT * ntrib, rel 3 rivT * rel * ntrib 1 rivT * ntrib, mdd 2 rivT * ntrib, mdd, rel 3 rivT * ntrib 4 rivT * ntrib, rel 5 rivT * rel * ntrib, mdd AICc 363.8 366.8 370.5 ∆AIC 0.0 3.0 6.7 wi 0.79 0.18 0.03 770.2 0.0 0.75 772.7 2.5 0.22 778.3 8.1 0.01 912.3 0.0 0.45 913.4 1.1 0.26 914.5 2.2 0.15 915.9 3.6 0.07 917.1 4.8 0.04 £ 1.00 0.22 0.04 1.00 0.29 0.02 1.00 0.58 0.33 0.17 0.09 Indian River Morning Midday (release) Afternoon 1 ntrib 2 ntrib, rel 3 ntrib, rivT 4 ntrib, mdd 1 rivT * rel, ntrib 2 rivT, rel, ntrib 3 ntrib, rel 1 ntrib 2 ntrib, rivT 3 ntrib, rel 418.0 0.0 0.75 1.00 421.0 3.0 0.17 0.22 423.7 5.7 0.04 0.06 424.8 6.8 0.03 0.03 534.0 0.0 0.97 1.00 542.9 8.9 0.01 0.01 544.1 10.1 0.01 < 0.01 444.8 0.0 0.72 1.00 448.2 3.4 0.13 0.18 448.3 3.5 0.12 0.17 Hudson River Midday 1 rivT * ntrib, nlog 360.6 0.0 0.43 1.00 2 rivT * ntrib, rel, nlog 361.0 0.4 0.35 0.82 3 rivT * rel * ntrib, nlog 363.1 2.5 0.12 0.29 4 rivT * ntrib, mdd, rel, nlog 365.3 4.7 0.04 0.10 5 rivT * ntrib, mdd, nlog 365.4 4.8 0.04 0.09 6 rivT * rel * ntrib, mdd, nlog 366.9 6.3 0.02 0.04 Afternoon (release) 1 rivT * rel * ntrib, mdd, nlog 561.9 0.0 0.83 1.00 2 rivT * rel * ntrib, nlog 565.8 3.9 0.12 0.14 3 ntrib * rel, mdd, rivT, nlog 569.6 7.7 0.02 0.02 29 Table 1.8. Predictor weights (w+(j)) for the three most important fixed effects based on AIC analysis for each river / time-of-day combination (where rivT = river temperature, ntrib = nearness to a tributary, rel = release day, mdd = mean daily discharge). Increasing values represent greater importance. The time periods during which the recreational release pulse passed the Indian and Hudson River reaches are highlighted with grey and parameters that include release day are bold. Morning Midday Afternoon CR rivT * ntrib = 0.97 rel = 0.18 rivT * ntrib * rel = 0.03 rivT * ntrib = 0.98 rel = 0.22 Mdd = 0.02 rivT * ntrib = 0.93 mdd = 0.76 rel = 0.34 IR ntrib = 1.000 rel = 0.18 rivT = 0.06 ntrib = 0.98 rivT * rel = 0.98 rel = 0.02 ntrib = 1.000 rivT = 0.15 rel = 0.15 HR not enough data rivT * ntrib = 0.86 rel = 0.39 rivT * ntrib * rel = 0.14 rivT * ntrib * rel = 0.94 mdd = 0.88 rivT = 0.04 (18°C; t180 = 2.90, p < 0.01) and high (22°C; t180 = -6.19, p < 0.01) values (Table 1.10), indicating that behavioral thermoregulation was more prevalent for trout located near tributaries. When river temperature was within the upper range of observed values, TD of trout near a tributary were more negative (LSMean = -1.49 ± 0.20°C) on average than those farther from tributaries (LSMean = -0.28 ± 0.16°C) (Figure 1.6). Conversely, when river temperature was held constant at the lower value in the model, observations more than 50 meters from a tributary were more negative. Although release day (i.e. whether or not a release occurred on a given day) appeared in both the second and third models, the main effect was not significant in the second model (release day, F1,167 = 0 .02, p > 0.50) and no comparisons of least-square mean estimates made between release and non-release days were significant for the third model (release day * near tributary * river temperature, F1,164 = 4.35, p = 0.04) (Figure 1.7). Additionally, the interaction between river temperature and nearness to a 30 tributary was, by far, the most important parameter (w+(j) = 0.97) with a relative importance five-fold greater than release day (Table 1.8). Results during midday in the Cedar River reach were very similar to the morning. The same model (river temperature * nearness to tributary) had the greatest support (over three times more support than the next best model) during both time periods (Table 1.7), and the fixed effect explained a significant amount of the variability (F1,287 = 14.27, p < 0.01) (Table 1.9). The differences between least-square mean estimates of observations within and farther than 50 meters from a tributary were significant only when river temperature was within the upper range of observed values (held constant at 22°C; t281= -4.34, p < 0.01) (Table 1.10). Under these conditions TD was more negative when trout were near a tributary (LSMean = -1.12 ± 0.18°C) than not (LSMean = -0.38 ± 0.15°C) (Figure 1.6). The second and third best models were also the same as during the morning, with release day appearing in both, but not statistically significant (model 2: release day, F1,21 = 0.13, p > 0.50; model 3: release day * near tributary * river temperature, F1,281 = 0.15, p > 0.50) (Table 1.7). The interaction between river temperature and nearness to a tributary had the largest predictor weight (w+(j) = 0.98) and had almost five times more weight than release day (w+(j) = 0.22) (Table 1.8). The best model during the afternoon for the Cedar River included the interaction between river temperature and nearness to a tributary and the mean daily discharge (Table 1.7). This model was nearly twice as well supported as the next best model. Both prediction variables explained a significant amount of the variation in the TD (mean daily dishcarge, F1,18.7 = 13.32, p < 0.01; river temperature * near tributary, F = 12.29, p < 0.01) (Table 1.9). The difference between the least-square mean estimates was significant when river temperature was within the upper range of observed values (held constant at 24°C; t279 = -5.78, p <0.01) (Table 1.10). 31 Table 1.9. Covariance parameters for the random effects (individual study fish = fishID, individual days = day, unexplained error = residual) and test for significance of fixed effects (ambient river temperature = rivT, nearness to a tributary = ntrib, mean daily discharge = mdd, release day = rel, “*” = interaction term) for the reported multilevel models for all Cedar River / time-of-day combinations. The unconditional model includes only random effects. Significance of included fixed parameters are reported where “ns” indicates not significant (α=0.05), “*” indicates significance at α=0.05 and “**” indicates significance at α=0.01. 32 Cedar Covariance Parameters Test for fixed effects River Model fishID day residual rivT ntrib mdd rel rivT * ntrib rivT * rel ntrib * rel rivT * ntrib * rel Morning unconditional 0.32 1 0.24 0.07 0.00 0.43 0.33 ** ** ** 2 0.26 0.00 0.33 ** ** ns ** 3 0.26 0.00 0.33 ** ** ns ** ns * * Midday unconditional 0.27 1 0.19 0.04 0.02 0.69 0.64 ** ** ** 2 0.21 0.02 0.64 ** ** ns ** 3 0.21 0.02 0.65 ** ** ns ** ns ns ns Afternoon unconditional 0.97 1 0.50 0.20 0.06 1.24 1.02 ** ** ** ** 2 0.52 0.07 1.01 ** ** ** ns ** 3 0.52 0.15 1.01 ** ** ** 4 0.52 0.16 1.01 ** ** ns ** 5 0.51 0.07 1.01 ** ** ** ns ** ns ns ns Table 1.10. Comparisons of least-square mean (LSMean) estimates of the temperature difference between fish body and ambient river for the most parsimonious multilevel model for each Cedar River / time-of-day combination. The left and right panels show results when holding river temperature constant within the lower (18˚C or 19˚C) and upper (22˚C or 24˚C) range of the data, respectively. Significance level of LSMean differences for observations <50 and >50 meters from a tributary are reported in the rightmost column of each panel. time period morning midday afternoon > 50m from < 50 m from tributary tributary LSMean (±SE) LSMean (±SE) Pr > F ambient river temperature = 18°C -0.06 ± 0.16°C 0.44 ± 0.20°C < 0.01 -0.05 ± 0.16°C 0.34 ± 0.25°C 0.12 ambient river temperature = 19°C -0.33 ± 0.24°C -0.32 ± 0.34°C > 0.50 > 50m from < 50 m from tributary tributary LSMean (±SE) LSMean (±SE) Pr > F ambient river temperature = 22°C -0.28 ± 0.16°C -1.49 ± 0.20°C <0.01 -0.38 ± 0.15°C -1.12 ± 0.18°C <0.01 ambient river temperature = 24°C -0.71 ± 0.23°C -1.96 ± 0.26°C <0.01 Under these conditions, TD was more negative when trout were near a tributary (LSMean = -1.96 ± 0.26°C) than when not (LSMean = -0.71 ± 0.23°C) (Figure 1.6). The magnitude of the temperature difference was greatest during the afternoon time period. With all other variables held constant in the model, the temperature difference increased at a rate of 0.02 ± 0.01°C for every cubic meter per second increase in mean daily discharge. The second, fourth and fifth best models all included release day, but release day did not explain a significant amount of the variability in the TD in any of these models (Model 2: release day, F1,17 = 0.71, p = 0.41; Model 4: release day, F1,20.4 = 0.02, p > 0.50; Model 5: release day * near tributary * river temperature, F1,272 = 0.38, p > 0.50) (Table 1.7). The interaction between river temperature and nearness to a tributary had the largest predictor weight (w+(j) = 0.93). The second most important factor was mean daily discharge (w+(j) = 0.76) which was more than twice as important as release day (w+(j) = 0.34) (Table 1.8). 33 Estimated TD (°C) Estimated TD (°C) 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Cedar River Morning -2.5 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Cedar River Midday -2.5 river temperature = 18 °C p < 0.01 river temperature = 18 °C p = 0.25 1.5 1 river temperature = 19 °C 0.5 p > 0.50 0 -0.5 -1 -1.5 -2 Cedar River Afternoon -2.5 > 50 m from tributary < 50 m from tributary Estimated TD (°C) Estimated TD (°C) Estimated TD (°C) 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Cedar River -2.5 Morning river temperature = 22 °C p < 0.01 1.5 1 river temperature = 22 °C 0.5 p < 0.01 0 -0.5 -1 -1.5 -2 Cedar River Midday -2.5 1.5 1 river temperature = 24 °C 0.5 p < 0.01 0 -0.5 -1 -1.5 -2 Cedar River -2.5 Afternoon > 50 m from tributary < 50 m from tributary Estimated TD (°C) Figure 1.6. Comparison of least-square mean (LSMean) estimates of the difference between fish body and river temperature from the most parsimonious multilevel model for each Cedar River / time-of-day combination. Plots on the left and right panels show results when holding river temperature constant in the model within the lower (18˚C or 19˚C) and upper (22˚C or 24˚C) range of the data, respectively. Results from morning, midday and afternoon are shown in the top, middle and bottom panels, respectively. Significance of the difference between the LSMean estimates for study trout < 50 meters or >50 meters from a tributary are reported. In the morning in the Indian River, the time period before the release pulse passed the reach, the best model included only one fixed effect, nearness to a tributary (Table 1.7). This model had nearly five times more support than the next best model. Nearness to tributary had a significant effect on the TD (F 1,231 = -5.61, p = 0.02) (Table 1.11) such that the differences between the least-square mean estimates showed that the body temperature of trout near a tributary were (LSMean = -0.19 ± 0.17 °C) 34 cooler relative to the river than those that were not near a tributary (LSMean = 0.21 ± 0.06°C; t231 = -2.37, p = 0.02) (Table 1.12 and Figure 1.7). Estimates of TD greater than zero can either be explained by fish being within patches of water warmer than the ambient river temperature (e.g. an unstratified pool with a smaller range in daily temperature than the ambient river) or by measurement error which could have originated from the accuracy of the radio transmitters and river temperature loggers or the distance and direction of the fish from the nearest river temperature logger. Release day was included in the second best model (Table 1.7), but did not explain a significant amount of the variation in TD (release day, F1,220 = 0.22, p > 0.50) (Table 1.11). Despite being included in fewer models than the other factors, nearness to a tributary had, by far, the largest predictor weight (w+(j) = 1.00) – more than five times as important as release day (Table 1.8). The best model during the midday time period in the Indian River included two fixed effects (Table 1.7), each of which explained a significant amount of the variation in the temperature difference: the interaction between river temperature and release day (F1,273 = 12.07, p < 0.01) and nearness to a tributary (F1,282 = 8.43, p < 0.01) (Table 1.11). This model was nearly 90 times as well supported as the next best model. Examination of the differences in least-square means showed a significant effect of release day on the TD only when river temperature was held constant in the model within the upper range of the data (24°C; t274 = -4.65, p < 0.01) (Table 1.13). Fish body temperature was estimated to be 0.24 ± 0.11°C less than the river on non-release days and 0.19 ± 0.11°C greater than the river on release days (Figure 1.8). The differences between least-square mean estimates of observations within and farther than 50 meters from a tributary were also significant (t282 = -2.90, p = 0.01) (Table 1.12). TD was more negative when trout were near a tributary (LSMean = -0.07 ± 0.15°C) than when not (LSMean = 0.34 ± 0.07°C) (Figure 1.7). Release day was also 35 Table 1.11. Covariance parameters for the random effects (individual study fish = fishID, individual days = day, unexplained error = residual) and test for significance of the fixed effects (ambient river temperature = rivT, nearness to a tributary = ntrib, mean daily discharge = mdd, release day = rel, “*” = interaction term) for the reported multilevel models for all Indian River / time-of-day combinations. The unconditional model includes only random effects. Significance of included fixed parameters are reported where “ns” indicates not significant (α=0.05), “*” indicates significance at α=0.05 and “**” indicates significance at α=0.01. 36 Indian River Morning Midday Afternoon Model unconditional 1 2 3 4 unconditional 1 2 3 unconditional 1 2 3 Covariance Parameters fishID day residual 0.04 0.01 0.29 0.03 0.00 0.31 0.03 0.00 0.31 0.03 0.00 0.31 0.03 0.00 0.30 0.09 0.06 0.27 0.09 0.00 0.27 0.08 0.01 0.27 0.06 0.02 0.28 0.19 0.00 0.27 0.18 0.00 0.25 0.19 0.00 0.25 0.18 0.00 0.25 rivT ns ** ** ns ntrib * * * * ** ** ** ** ** ** Test for fixed effects mdd rel rivT * ntrib ns ns ** ** ** ns rivT * rel ** Table 1.12. Comparison of least-square mean (LSMean) estimates of the difference between fish body and river temperature for the most parsimonious multilevel model for each Indian River / time-of-day combination. Significance level (corrected for multiple comparisons) of LSMean differences between observations of fish <50 meters and >50 meters from a tributary are reported time period morning midday afternoon > 50m from tributary LSMean (±SE) 0.21 ± 0.06 °C 0.34 ± 0.07 °C 0.37 ± 0.09 °C < 50m from tributary LSMean (±SE) -0.19 ± 0.17 °C -0.07 ± 0.15 °C -0.44 ± 0.19 °C Pr > F 0.02 0.01 <0.01 Table 1.13. Comparisons of least-square mean (LSMean) estimates of the difference between fish body and river temperature for the most parsimonious multilevel model during midday in the Indian River. Significance level of LSMean differences between release days and non-release days when holding values of ambient river temperature constant within the lower (20˚C) and upper (24˚C) range of the data are reported. time period midday ambient river temperature = 20°C non-release release LSMean (±SE) LSMean (±SE) Pr > F 0.40 ± 0.14°C 0.27 ± 0.12°C 0.45 ambient river temperature = 24°C non-release release LSMean (±SE) LSMean (±SE) Pr > F -0.24 ± 0.11°C 0.19 ± 0.10°C <0.01 included in the second and third best models, but these models had little support from the data (Table 1.7). Despite being included in fewer models than the other factors, nearness to a tributary was similarly important to the interaction between river temperature and release day (w+(j) = 0.98) (Table 1.8). Similar to the morning, during the afternoon time period in the Indian River the best model included only nearness to a tributary as a fixed effect and had more than five times as much support as the next best model (Table 1.7). Nearness to a tributary explained a significant amount of the variation in TD (nearness to tributary, F1,226 = 23.82, p < 0.01) (Table 1.11). Trout body temperatures were 0.44 ± 0.19°C 37 cooler than the river when observed within 50 meters of a tributary and 0.37 ± 0.09°C warmer than the river when more than 50 meters from a tributary (Figure 1.7), which was a significant difference (t226 = -4.84, p < 0.01) (Table 1.12). Although release day was included in the third best model (Table 1.7), this factor did not explain a significant amount of the variability in TD (release day, F1,241 = 0.10, p > 0.50) Estimated TD (°C) Estimated TD (°C) 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Indian River Morning -2.5 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Indian River Midday -2.5 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Indian River Afternoon -2.5 > 50 m from tributary p = 0.02 p = 0.01 p < 0.01 < 50 m from tributary Estimated TD (°C) Figure 1.7. Comparison of least-square mean (LSMean) estimates of the difference between fish body and ambient river temperature for the most parsimonious multilevel model for the Indian River during morning, midday and afternoon (top, middle and bottom panels, respectively). Significance values of LSMean estimates between observations of trout < 50 meters and > 50 meters from a tributary are reported. 38 (Table 1.11). As with the earlier time periods, nearness to a tributary was the most important factor (w+(j) = 1.00) and was over six times more important than release day (w+(j) = 0.15) (Table 1.8). 1.5 1 0.5 0 -0.5 -1 -1.5 Indian River -2 Midday -2.5 river temperature = 20 °C p = 0.45 Estimated TD (°C) 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Indian River Midday -2.5 river temperature = 24 °C p < 0.01 Figure 1.8. Comparison of the least-square mean (LSMean) estimates of the difference between fish body and ambient river temperature from the most parsimonious multilevel model for the Indian River during midday. Plots on the left and right panels show results when holding river temperature constant in the model within the lower (20˚C) and upper (24˚C) range of the data, respectively. Significance of the difference between the LSMean estimates for release days and non-release days are reported (filled squares represents non-release days and open squares represent release days). Sufficient data were only available during the midday and afternoon time periods for Hudson River analyses. During midday, the time period before the release pulse passed the study reach, the best model of the set included the following two fixed effects: the interaction of river temperature and nearness to a tributary (F1,119 = 18.78, p < 0.01) and distance from nearest logger (F1,111 = 5.28, p = 0.02) (Table 1.14). Including the distance of the fish from the nearest temperature logger accounted for variability due to more spatially infrequent sampling of ambient river temperature in the Hudson River compared to the other study reaches. This model had just slightly more support than the next best model and more than three times the support of the third best model (Table 1.7). The difference between the least-square mean estimates 39 Table 1.14. Covariance parameters for the random effects (individual study fish = fishID, individual days = day, unexplained error = residual) and test for significance of fixed effects (ambient river temperature = rivT, nearness to a tributary = ntrib, mean daily discharge = mdd, release day = rel, distance from nearest logger = nlog, “*” = interaction term) for the reported multilevel models for both Hudson River / time-of-day combinations. The unconditional model includes only random effects. Significance of included fixed parameters are reported where “ns” indicates not significant (α=0.05), “*” indicates significance at α=0.05 and “**” indicates significance at α=0.01. 40 Hudson Covariance Parameters Test for fixed effects River Midday Model fishID day residual rivT ntrib mdd rel nlog rivT * ntrib rivT * rel ntrib * rel rivT * ntrib * rel unconditional 0.20 0.79 0.63 1 0.12 0.89 0.34 ** ** * ** 2 0.12 0.40 0.34 ** ** ns * ** 3 0.12 0.43 0.34 ** ** ns * ** ns ns ns 4 0.12 0.36 0.34 ** ** ns ns * ** 5 0.12 0.35 0.34 ** ** ns * ** 6 0.11 0.37 0.34 ** ** * ns * ** ns ns ns Afternoon unconditional 0.42 0.53 1 0.28 0.22 0.96 0.64 ** ns ** ns ns * ns ** ** 2 0.34 0.27 0.65 ** * ns ns ** ns ** ** 3 0.30 0.29 0.67 ** ** ** ns ns ns Table 1.15. Comparisons of least-square mean (LSMean) estimates of the temperature difference between fish body and ambient river for the most parsimonious multilevel model for both Hudson River / time-of day-combinations. Significance of LSMean differences between fish within and farther than 50 meters from a tributary at values of ambient river temperature within the lower and upper range of the data are reported in the rightmost column of each panel. Significance of LSMean differences between release days and non-release days are reported in the bottom row for midday only. time period midday afternoon non-release release > 50m from < 50m from tributary tributary LSMean (±SE) LSMean (±SE) Pr > F ambient river temperature = 20°C -0.21 ± 0.24 °C -0.18 ± 0.27 °C > 0.50 0.98 ± 0.41 °C 1.26 ± 0.44 °C > 0.50 1.06 ± 0.32 °C 0.30 ± 0.32 °C > 0.50 Pr > F > 0.50 Pr > F > 0.50 > 50m from < 50m from tributary tributary LSMean (±SE) LSMean (±SE) Pr > F ambient river temperature = 25°C -0.76 ± 0.18 °C -2.12 ± 0.26 °C < 0.01 -0.21 ± 0.28 °C -2.04 ± 0.34 °C < 0.01 -0.56 ± 0.21 °C -0.89 ± 0.29 °C > 0.50 Pr > F > 0.50 Pr > F = 0.05 of TD for observations within 50 meters of a tributary (-2.12 ± 0.26°C) and more than 50 meters from a tributary (-0.76 ± 0.18°C) were significant (t128 = -5.85, p < 0.01) only when the river temperature was within the upper range of the data (held constant at 25°C) (Table 1.15 and Figure 1.9). Although release day was included in the second, third, fourth and sixth best models, this factor did not explain a significant amount of the variation in TD (Model 2: release day, F1,28.8 = 0.46, p = 0.50; Model 3: river temperature * nearness to a tributary * release day, F1,110 = 2.39, p = 0.12; Model 4: release day, F1,27.1 = 1.01, p = 0.32; Model 6: river temperature * nearness to a tributary * release day, F1,110 = 2.83, p = 0.10) (Table 1.7). The interaction between river temperature and nearness to a tributary was the most important variable with parameter weight equal to 0.86 and was more than twice as important as release day (w+(j) = 0.39) (Table 1.8). During the afternoon time period when the release pulse passed through the Hudson River reach, the best model included the following fixed effects: the interaction between river temperature, nearness to a tributary and release day (F1,143 = 11.66, p < 0.01), mean daily discharge (F1,27.5 = 12.12, p< 0.01), and distance from 41 nearest logger (F1,168 = 1.14, p = 0.29) (Table 1.14). This model had seven times more support from the data than the next best model (Table 1.7). Least-square mean estimates differed depending on whether a fish was near a tributary and on the occurrence of a pulsed discharge event, but only when river temperature was held constant in the model within the upper range of the data. Release day was only significant when observations were near a tributary and the river temperature was held constant in the model at a value within the upper range of the data (t69.9 = -2.79, Estimated TD (°C) 1.5 1 river temperature = 20 °C 0.5 p > 0.50 0 -0.5 -1 -1.5 -2 Hudson River Midday -2.5 1.5 1 0.5 P > 0.50 0 -0.5 p > 0.50 -1 -1.5 -2 Hudson River Afternoon river temperature = 20 °C -2.5 > 50 m from < 50 m from tributary tributary Estimated TD (°C) Estimated TD (°C) 1.5 1 0.5 0 -0.5 -1 -1.5 -2 Hudson River Midday -2.5 river temperature = 25 °C p < 0.01 1.5 1 river temperature = 25 °C 0.5 p > 0.50 0 -0.5 p = 0.05 -1 -1.5 -2 Hudson River Afternoon -2.5 > 50 m from tributary < 50 m from tributary Estimated TD (°C) Figure 1.9. Comparisons of least-square mean (LSMean) estimates of the temperature difference between fish body and ambient river for the most parsimonious multilevel model for the Hudson River during midday and afternoon time periods (top and bottom panels, respectively). Results when river temperature was held constant at a value within the lower (20˚C) and upper (25˚C) range of the data are shown in the left and right panels, respectively. For midday panels, significance of the LSMean for trout <50 and >50 meters from a tributary are reported. For the afternoon panels, the significance of two LSMean comparisons are reported: 1) the difference between release and non-release days for observations > 50 meters from a tributary and 2) the difference between release and non-release days for observations < 50 meters from a tributary. Filled squares represents non-release days and open squares represent release days. 42 p = 0.05) (Table 1.15), where the least-square mean estimate of the TD was -2.04 ± 0.34°C on non-release days and -0.89 ± 0.29°C on release days (Figure 1.9). Similarly, significant differences existed between trout near tributaries only in the absence of releases and at ambient river temperatures within the upper critical range (t169 = -4.93, p < 0.01) (Table 1.15).With all other variables held constant, the temperature difference increased at a rate of 0.03 ± 0.01°C for every cubic meter per second decrease in mean daily discharge. Release day was also included in the second and third best models (Table 1.7). The interaction between river temperature, nearness to a tributary and release day was the most important variable with parameter weight equal to 0.94. Mean daily discharge was also important (w+(j) = 0.88) (Table 1.8). Discussion This study demonstrated that recreational discharge events in the Indian and Hudson Rivers increased both the daily temperature maximum and range within thermal refuge areas and were associated with a reduction in behavioral thermoregulation by stocked brown trout. Almost no study trout persisted within these river reaches for an entire summer, while approximately half of the study trout in a reference reach unaffected by releases (Cedar River) survived through late August. Although indirect effects of recreational flow releases likely contributed to mortality, the thermal regimes within both the Indian and Hudson River reaches, regardless of the pulsed discharge events, were thermally marginal and less hospitable for brown trout than the Cedar River reach. 43 River temperature Based on both laboratory-derived tolerance values and field-based thresholds that incorporated metrics of temperature magnitude, duration, and fluctuation, all three rivers were thermally marginal for brown trout during the summers of 2005 and 2006. The summer of 2005 had higher temperatures than the summer of 2006. During the 2005 study period, both the Indian and Hudson Rivers were continuously within the range of temperatures where normal brown trout behavior is disrupted (Elliot 1994). The 2006 study period was milder, but study reaches still reached temperatures expected to limit brown trout presence. Temperature thresholds were exceeded to a greater extent at longer (3-week to 2 month) exposure periods than at short-term (1day to 7-day) exposures, suggesting that the sustained high summer temperatures would be more limiting for brown trout persistence in these rivers than short term temperature extremes. The thermal regime varied between the three study reaches. The Indian River had the most stable water temperature, which likely resulted from being just downstream from an impoundment and minimal tributary and groundwater input (Appendix B; Webb and Walling 1996). This reach also had warmer average daily temperatures than the other two study reaches. Similarly, both the Hudson River and the Cedar River reaches had greater daily temperature ranges than the Indian River, which may have provided resident trout a period of recovery from high daytime temperatures, i.e. during nighttime minimums (Johnstone and Rahel 2003; Wehrly et al. 2007). Given that average daily temperatures exceeded those expected to support brown trout for long exposure periods, it is likely that these temperature fluctuations may have been beneficial in early summer, but later became a source of additional stress (Jobling 1997; Wehrly et al. 2007). The Cedar River reach had the lowest average daily temperatures and thereby the most suitable thermal habitat of the study 44 reaches. Considering the marginal conditions in our three study reaches, especially in the Hudson and Indian Rivers, brown trout could only be expected to survive within reaches where thermal refugia were available. Effects of recreational discharge events on thermal refuge habitat While recreational discharge events did not significantly increase the mean or maximum daily temperature in either the mainstem Indian or Hudson River reaches (Baldigo et al. in prep), the temperature regimes within localized patches of cool water near tributary confluences (where mixing of cool tributary water with warmer mainstem water occurs) were diluted by release events. Both daily temperature maxima and ranges increased significantly on days with recreational flow releases. The magnitude of this increase was greater at low base flow and varied between the three monitored tributaries to the Hudson River. No relationship was found between the occurrence of releases and the average daily temperature. This may have been due to the short period of increased temperature associated with a discharge event (3-5 hours) that was subsequently countered by a decrease in thermal refuge dilution (and temperature) associated with a post-release drop in discharge as the upstream reservoir recharged. Acute or chronic stress to fish can be caused by sudden temperature changes and fluctuating temperatures at high average values (Flodmark et al. 2002; Flodmark et al. 2004; Quigley and Hinch 2007; Wehrly et al. 2007). When low base flows coincide with warm summer temperatures, patches of cool water provide a reduction in daily temperature maxima, average and range. Conversely, when diluted by recreational flow releases these same thermal refugia may become zones of rapid temperature increases and larger daily temperature fluctuations than would occur in the mainstem river. Although some trout were observed moving farther upstream into 45 tributaries as the release pulse passed – thereby avoiding refuge dilution – this movement was impossible when tributary flows were at summer lows. Differences in tributary morphology can contribute to different thermal conditions in associated thermal refuge areas (Nielsen et al. 1994). The ability of large fish to move upstream into both the Balm of Gilead and Griffin Brook was reduced or blocked by exposed cobble bars during periods of low summer base flow. Under these conditions, cold water from these tributaries seeped into the shallow interstices along the river edge, but cover and sufficient depth suitable for large fish were generally unavailable. At the mouth of Raquette Brook, a shallow pool with overhanging vegetation and large boulders suitable for cover was available throughout the 2006 study period. However, lower flow conditions in 2005 substantially decreased the availability of this habitat at this location. Although we observed dilution from recreational flow releases in the Raquette Brook confluence, the magnitude of the disturbance was more moderate than at the other two tributaries. Effects of recreational discharge events on trout behavioral thermoregulation Behavioral thermoregulation was observed in the adult brown trout in our study in all river reaches, although infrequently in the Indian River. We found that trout in the Cedar River were more often observed with body temperatures cooler than ambient river temperature (38%) than those in either the Hudson (29%) or Indian Rivers (4%). These observed proportions are low but within the range of those reported from other investigations of salmonid use of thermal refugia. For example, Ebersole et al. (2001) found that the proportion of rainbow trout within thermal refuge areas ranged from 10% to 40% in northeastern Oregon streams. Nielsen et al. (1994) found that 65% of steelhead trout moved into stratified pools during midday or afternoon in a northern California stream that reached temperatures as high as 28°C. 46 As expected, transmitter-implanted trout in our study exploited thermal refugia, but the relatively low rate of use suggests that these areas of cool water were limited. Considering the thermally stressful conditions in all study reaches and the poor persistence reported for stocked brown and rainbow trout in other studies (Skurdal et al. 1989; Bettinger and Bettoli 2002; Pedersen et al. 2003), it is not surprising that fewer than 50% of stocked fish persisted over a 67 day period in all three of our study reaches. Persistence of stocked brown trout in the Cedar River in 2006 was greater than in either the Hudson or Indian Rivers in either 2005 or 2006. This result was likely due, in part, to the more suitable thermal regime and greater amount of thermal refugia associated with cover found in the Cedar River reach (Appendix B), but may also be related to increased thermal stress caused by the occurrence of recreational flow releases in the Indian and Hudson River reaches. The greater mortality rate observed during 2005 for both the Indian and Hudson Rivers was likely due to warmer temperatures during that year and possibly to the fact that the fish were stocked in mid-July when river temperatures were already stressful. Because recreational flow releases were not an important factor accounting for the thermal behavior of trout during any time period in the reference Cedar River reach (without dam releases), we conclude that the study design successfully captured the effects of environmental variables, while sampling without bias for days designated as release or non-release. Furthermore, the occurrence of recreational flow releases was not an important factor influencing behavioral thermoregulation in the two affected reaches during time periods when the release pulse was not present within that reach. This validates our findings that within the Indian and Hudson River reaches, the occurrence of recreational flow releases was an important factor determining behavioral thermoregulation during midday and afternoon time periods, respectively. 47 During the time periods when the recreational discharge pulse passed through the affected study reaches (midday for the Indian River and afternoon for the Hudson), release day was a relatively important factor influencing behavioral thermoregulation of brown trout. In both reaches, behavioral thermoregulation was reduced when inundated by the release pulse. This reduction occurred only when the river temperatures were within the upper critical range for brown trout and, for the Hudson River, when study trout were near tributary confluences. Likely due to the paucity of thermal refuge areas and the few observations of trout in thermal refugia that could be disturbed by releases, the magnitude of the release day effect in the Indian River reach was less than 0.5°C and was therefore not likely to be biologically significant. In the absence of recreational discharge events, the most important factors affecting behavioral thermoregulation were whether a study trout was located near a tributary confluence and the ambient river temperature. Brown trout near tributaries were consistently cooler relative to ambient river temperature than those located more than 50 meters from a tributary confluence when river temperature was within the upper critical range. Although other sources of cold water refuge were available, this finding suggests that tributaries were an important thermal resource. The temperature differences were the most negative during the afternoon peak in water temperature, demonstrating that trout increased behavioral thermoregulation as river thermal conditions became more stressful. Observed temperature differences in Hudson and Cedar River study fish were more negative (i.e. the fish were cooler than the river) than Indian River study trout in all time period comparisons. During the afternoon, mean daily discharge was an important factor explaining the variation in brown trout behavioral thermoregulation in both the Cedar and Hudson River reaches; however, mean daily discharge had opposite effects within the two reaches. As mean daily discharge (a surrogate for base flow) decreased, the 48 temperature difference between fish body and ambient river became more negative (indicating increased behavioral thermoregulation) in the Cedar River reach and less negative in the Hudson River reach. The trend in the Cedar River reach, equivalent to the reduction in temperature at low base flow within our monitored tributary confluences, was likely due to decreased influence of mainstem river water in refuge areas or decreased flows reducing mixing of other cold water patches (Nielsen et al. 1994; Matthews et al. 1994). The opposite trend – decreases in behavioral thermoregulation at lower base flows – was observed for trout in the Hudson River reach. This trend was not associated with recreational releases and was likely due to physical characteristics of the available cool-water habitats. The quality of thermal refugia at tributary confluences was influenced by the flows from and geomorphic structure of both the tributary and the mainstem channel. Most of the thermal refuge areas used by transmitter-implanted trout in the Hudson River were within or at the confluence of adventitious streams (low order tributaries feeding higher order rivers). The greater difference in stream order between the Hudson River and its cool tributaries compared with a smaller size difference for tributaries to the Cedar and Indian Rivers – coupled with the wide, shallow morphology (Baldigo et al. in prep) of the mainstem channel – may have contributed to the ephemeral nature of thermal refugia in the Hudson River study reach and, in turn, the decrease in behavioral thermoregulation by trout at low base flows. Although we did observe cooler temperatures within the confluences of representative adventitious streams in the Hudson River at low base flows, the lack of cover likely rendered them unsuitable habitat. During low summer flows the mainstem channel retreated from any overhanging vegetation and became shallow within the confluences of Griffin Brook, Balm of Gilead and other similar areas. These locations, which had provided thermal refugia earlier in the season, likely became either uninhabitable or 49 areas of high predation risk due to a decline in the quantity or quality of the associated physical habitat. In 2006, the Raquette brook confluence was the monitored thermal refuge area least affected by releases, was one of the few Hudson River cool-water areas that was associated with sufficient depth and cover, and was the most common observed location of fish in the Hudson River reach (Appendix B). In the Cedar River reach, deep stratified pools and runs and tributaries that were often associated with undercut banks and overhanging vegetation were available throughout the summer (Appendix B). Angling during summer 2006 revealed abundant brook trout within the mainstem and cold tributaries of the Cedar River between the Wakely Dam and the Cedar River Dam, suggesting the capacity of the Cedar River to sustain populations of coldwater fishes. Although adventitious tributaries can provide thermal relief for salmonids (Thomas and Hayes 2006), the size and quality of associated physical habitat characteristics of those in the Hudson River reach may have been insufficient to sustain adult brown trout during the critical summer months when low base flow coincided with high summer temperatures, whereas in the Cedar River, where more than 50% of monitored brown trout survived the summer, the variety and overall quality of thermal refuge habitat appeared sufficient even at low flows. The small number of brown trout observed to behaviorally thermoregulate in the Indian River was likely due, in part, to a lack of available thermal refugia. Only three tributaries entered this reach, with only one accessible to large brown trout throughout the summer. Considering this and the fact that only 40% of fish observations occurred in cool water patches in any reach suggest that thermal refugia were limited and / or that other locations provided more suitable habitat. Observations of monitored trout in this study suggest that aspects of the stream environment other than temperature may have influenced behavior. Other research has 50 found that salmonids do not select position based solely on optimal temperature, but rather select habitat with favorable physical attributes that occur within a thermal tolerance range (Spigarelli et al. 1983; Matthews et al. 1994; Neilsen et al. 1994). Transmitter-implanted trout in all reaches of this study were observed in plunge pools, despite the lack of thermal relief in many of these areas. This was particularly true in the Indian and Cedar River study reaches where the majority of observed fish locations were within such habitat. Additionally, most monitored brown trout in both of these reaches were frequently found within 1 km of a dam and may have benefited from an increased influx of prey derived from the impoundments (Appendix B). When faced with greater metabolic demand associated with high temperatures, brown trout in these rivers may not only be seeking cooler water, but may be maximizing food intake and selecting slow water areas. The presence of deep slow habitat and abundant food may have influenced fish in the Indian and Cedar Rivers to select such areas over cold water refugia. In thermally marginal streams such as those that we studied in the Upper Hudson River drainage, accessible thermal refuge areas are important resources that provide trout a haven from lethal summer temperature conditions. Although important to survival, these areas are limited and most restricted during low summer flows. When low flow conditions correspond with peak summer temperatures, these refuge areas are likely most important and most vulnerable to altered flow regimes. Our results showed that pulsed discharge events alter both the thermal characteristics of refuge areas at tributary confluences and behavioral thermoregulation by stocked brown trout within the affected reaches, one of which was 30 kilometers downstream from the release source. Although poor survival of stocked brown trout in these affected reaches may be due to summer temperatures that exceed established tolerances regardless of recreational releases, the observed reduction in behavioral 51 thermoregulation during pulsed discharge events suggests that they may impair the ability of coldwater fish to survive in regulated river systems. Implications Results from this research effort have important implications for management of salmonid species in the studied reaches of the Upper Hudson River drainage, as well as coldwater fisheries management throughout New York State. Management options are constrained by both the physical characteristics of a particular river system and the legal framework and administrative authority pertaining to a specific location. This was particularly evident with regard to the subject study area of this thesis, where all the study reaches were designated by New York State Department of Environmental Conservation as wild, scenic or recreational and portions of the Hudson River were encompassed by the Hudson Gorge Primitive Area. Each designation imparted a specified level of protection from alteration, development and use. Fisheries management goals and decisions must be made within this context. Results from this study show that warm summer temperatures in both the Indian River below Lake Abanakee and the Hudson River near the hamlet of North River, NY make it very unlikely that a successful holdover brown trout fishery could be sustained under current climate conditions. Major modifications to the river morphology, such as increasing depth within or adding large woody debris to thermal refuge areas, might enhance thermal refuge habitat but are costly and would likely face legal restrictions in these protected river reaches. Maintaining a seasonal put-andtake fishery is a more realistic goal. Given the more sedentary behavior and longer persistence of brown trout in the Indian River study reach, ongoing stocking efforts at that location will likely continue to produce a more successful spring and summer fishery than in the Hudson River near North River. 52 The actual success of the current stocking program or any future adjustments could be evaluated with a creel survey. Information regarding the current catch rate of stocked fish, as well as other species targeted by anglers would provide a more informed knowledge base for management decisions. To more definitively determine the presence or absence of holdover salmonids within these reaches, marking (e.g. fin clipping) all stocked fish (including those privately stocked) would be crucial. Brook trout were observed (via angling or visual identification in shallow thermal refugia) in both the Indian and Hudson Rivers during the study, but without any identifying marks for stocked fish it was impossible to determine whether these fish originated from natural reproduction or stocking in private waters connected to the study reaches. If self-sustaining populations of native brook trout exist within these reaches, competition for limited resources between these and stocked fish should be considered and the possibly conflicting management goals of preserving native populations and providing a recreational fishery should be evaluated. Additionally, information on the distribution of self-sustaining coldwater species throughout New York state rivers, combined with current or expanded temperature monitoring could be developed into guidelines that describe New York specific salmonid temperature tolerances – following the approach of Wehrly et al. (2007) that evaluated the distribution of both salmonine fish populations and river temperature metrics. Finally, pulsed discharge events, although not likely the ultimate cause of poor survival in this study, appear to have a negative impact on adult brown trout behavior. During late July and early August, peak temperatures and lowest flows coincide with periods of prolonged thermal stress. This is also the time when thermal refuge areas are most susceptible to dilution, therefore any thermal habitat mitigation efforts might best be focused during this time period. These findings are not immediately applicable to other river systems because local differences in physical habitat, thermal refuge 53 areas, community composition, and species of management priority determine the scope of potential problems and possible solutions for such regulated flows. Nonetheless, the impacts of pulsed discharge events on thermal refugia and salmonid behavior should be recognized as a potential problem and investigated in other systems to create a larger body of knowledge on this topic. 54 APPENDIX A FATE OF STOCKED 2-YEAR-OLD BROWN TROUT IN THE UPPER HUDSON RIVER DRAINAGE While mortality due to acute or accumulated thermal stress likely contributed to the short persistence time of brown trout in our study, other direct and indirect causes of mortality were also responsible for the loss of fish from study reaches. The fate of fish – either being lost from the study due to emigration or mortality, or remaining alive and within the study reach (Bettinger and Bettoli 2002) – can provide insights into likely causes of mortality within a study area. Previous investigations attributed mortality of stocked brown trout in streams to direct causes including angling (Cresswell 1981; Skurdal et al. 1989; Aarestrup et al. 2005; Baird et al. 2006), predation by mink (Mustela vison), otter (Lutra lutra), osprey (Pandion haliaetus) and heron (Ardea cinerea) (Pedersen et al. 2003; Diana et al. 2004; Lindstrom and Hubert 2004; Aarestrup et al. 2005) and stranding during rapid dewatering (Saltveit et al. 2001). Others have suggested indirect causes as contributors to mortality, such as the poor ability of stocked salmonids to minimize energetic costs, find food or avoid predation (Bachman 1984; Bettinger and Bettoli 2002; Aarestrup et al. 2005), the physiological stress of stocking (Skurdal et al. 1989) and high levels of activity (Aarestrup et al. 2005; Scruton et al. 2005). We estimated the fate of study fish based on the confirmed or inferred final resting locations of each transmitter. We posted signs along the river banks of all study reaches to inform recreational users of the study and request the return of transmitters retrieved by anglers. In addition, throughout the summer we attempted to retrieve transmitters as soon as possible after receiving a mortality signal. This pursuit resulted in the finding of whole or pieces of dead fish, just the transmitter, or sometimes an 55 observation of a living fish with a radio transmitter. Transmitters were sometimes tracked to informative locations that included a heron and osprey rookery, guano beneath large pine trees, and burrows within stream banks or on the forest floor. Other final resting locations of transmitters were less conclusive, including locations within the main river channel and along the river margin in water depths not likely accessible to adult trout under base flow conditions. Approximately fifty percent of transmitters were recovered, and the remaining final resting locations were inferred (Appendix E). The fate of each fish was assigned to one of the following subcategories: ‘signal lost’ indicates that the signal for the transmitters ceased being detected during the study; ‘in woods’ describes locations beyond the width of the river at the highest summer flood (one fish in this category in 2006 was taken by an angler); ‘flood zone or shallow water’ describes locations with water depth not likely accessible to adult trout under base flow conditions, but flooded during recreational flow releases and naturally high discharges; ‘mid-channel’ describes locations within the river other than shallow water; and ‘in living fish’ refers to a transmitter that showed evidence (either by movement or visual observation) of remaining implanted in a living fish at the end of the study. It is possible that some of the transmitters in the ‘mid-channel’ category were expelled by living fish, in which case a small number of fish may have survived past the end date assigned by our criteria. Transmitter expulsion by brown trout has been observed in other studies (Jepsen 2008), and the rate of expulsion by rainbow trout has been shown to increase at warmer water temperatures (Bunnell and Isely 1999). It is possible that brown trout reared in a spring-fed (approximately 11°C) hatchery facility during this study may have exhibited a lower and slower rate of transmitter expulsion than fish from the same source that were released into the river because hatchery water temperatures were cooler. None of these hatchery-held fish expelled transmitters 56 during the time frame of the study. Additionally when whole dead fish were found in the field, there was no indication of imminent expulsion. Therefore we don’t believe that transmitter expulsion was common for our implanted trout. Signals were lost from approximately one third of deployed transmitters in the Hudson River reach during both 2005 and 2006 (Figure A.2), but in both the Indian and Cedar River reaches, nearly all transmitters were accounted for at the end of the study. A loss of signal could have resulted for a number of reasons. The most likely causes were unreported catch by anglers and loss from mammalian or avian predators that could carry the fish and transmitter beyond the range of signal detection. Rapid emigration from the river reach due to competitive displacement is another possibility that has been suggested for recently stocked brown trout (Popoff and Neumann 2005). 18 Hudson River 2006 16 Indian River 2006 Cedar River 2006 14 Hudson River 2005 12 Indian River 2005 10 # of Transmitters 8 6 4 2 0 Signal lost In woods Flood zone or Mid-channel In living shallow water fish Figure A.1. Confirmed or inferred locations of transmitters at the conclusion of the study during 2005 and 2006. 57 We attempted to manage the possibility that fish emigrated from the study reach by deploying a stationary receiver in 2006 approximately 10 km downstream from the Hudson River stocking location for the first 10 days post stocking. The receiver continuously scanned for each Hudson River transmitter every five minutes during this entire period, resulting in no observations of downstream emigration. This continuously monitoring receiver was subsequently placed upstream from the stocking location for the next three weeks, until July 18, 2006. The receiver did not provide any additional information regarding lost signals during this time, such as long distance nighttime movements that have been observed in previous studies (Clapp et al. 1990; Diana et al. 2004). At the confluence of the Indian River with the Hudson – approximately 20 km upstream from the Hudson River study reach – we regularly scanned for missing Hudson River study fish in both 2005 and 2006, but never detected a transmitter. Additionally, once each year we hired a rafting company to negotiate the gorge section of the Hudson River while we scanned for study trout, but again no signals were detected. Transmitters or dead fish observed or inferred to be in the woods were attributed to predation. Observations of transmitters in burrows and rookeries and of osprey catching large fish from both the Indian and Hudson River study reaches validated this conclusion. Additionally, we observed mammalian tracks thought to be mink and raccoon as well as scat containing fish scales at all three sites, and a mink family was encountered along the banks of the Hudson River site in 2006. It is also possible that fish that died within the river were subsequently removed by scavenging animals and carried into the surrounding wooded areas. The higher proportion of fish in the Indian and Hudson Rivers taken by predators may be a result of poor condition of fish in these reaches relative to the Cedar River or the more readily available cover within the Cedar River, in either case leaving trout in the Indian and Hudson reaches 58 more susceptible to predation. During post-study observations within the Cedar River reach in fall 2006, we found a transmitter from a trout that had survived throughout the study period buried in a stream bank, indicating the continued loss of fish to predation after the conclusion of the study. We inferred the cause of death for fish with transmitters found within the flood zone or shallow water to be due to either predation or stranding. Stranding has been documented for juvenile salmonids (Bradford 1997; Saltveit et al. 2001; Scruton et al. 2003), but not adults. Adults that use flood zone habitat during recreational releases (Bunt et al. 1999; Heggenes et al. 2007) may be susceptible to stranding during the post-release reduction in discharge. A heron was observed eating one of our implanted fish within the flood zone, lending credibility to the idea that predation was responsible for the occurrence of transmitters within the flood zone. Again the possibility exists that in-channel death followed by scavenging may have resulted in these shallow water final resting locations. Although other indirect causes likely contributed to the steady decline in the number of fish alive within our study reaches (Figure A.2), the large number of transmitters confirmed or inferred to be in the woods and exposed in the flood zone suggests that predation may have been an important cause of mortality in all three rivers. The high loss of fish suggests that survival for stocked brown trout within the Indian and Hudson Rivers is poor. Although a small number may survive and possibly overwinter, it is unlikely that a population of brown trout is holding over in these reaches. 59 16 14 12 10 # of fish still 8 in study 6 4 2 0 CR 2006 IR 2006 HR 2006 IR 2005 HR 2005 14Jun 19Jun 24Jun 29Jun 4Jul 9Jul 14Jul 19Jul 24Jul 29Jul 3Aug 8Aug 13Aug 18Aug Figure A.2. The number of implanted fish alive and within the study reach plotted for each day of the study in 2005 and 2006 for Indian and Hudson Rivers and 2006 for the Cedar River. Two additional transmitter-implanted fish were released into the Hudson River on July 10, 2006. 60 APPENDIX B MOVEMENT BEHAVIOR AND HABITAT USE OF STOCKED 2-YEAR-OLD BROWN TROUT IN THE UPPER HUDSON RIVER DRAINAGE Knowledge of the movement patterns and habitat use of stocked brown trout can provide insights into the variable survival of fish within our study reaches. Although poor survival is commonly observed for stocked catchable brown trout (Skurdal et al. 1989; Bettinger and Bettoli 2002; Pedersen et al. 2003), understanding potential indirect causes of mortality can lead to better informed stocking policies as well as an understanding of the natural variables and challenges that mediate survivorship in river reaches similar to those included in this study – notwithstanding the occurrence of pulsed discharge events. Previous studies have reported that hatchery origin catchable-size brown trout have typically been observed or recaptured close to the location where they were stocked (Cresswell 1981; Clapp et al. 1990; Diana et al. 2004; Popoff and Neumann 2005; Heggenes et al. 2007). For example, most brown trout were recaptured within 4.5 km of the stocking location in a number of studies summarized by Cresswell (1981). In a later investigation, most brown trout stocked into the Farmington River, Connecticut were observed within 500 meters and 930 meters after 2 and 12 weeks, respectively (Popoff and Neumann 2005). In an Adirondack river in New York, Baird et al. (2006) found that, on average, brown trout were caught within 500 meters of the stocking location. The summer activity patterns of large brown trout have generally been described as sedentary during the day and actively foraging either locally or over large distances during dusk, dawn and in some cases throughout the night, usually returning to one of several home locations in the morning (Clapp et al. 1990; Diana et al. 2004). 61 Heggenes et al. (2007), however, did not observe diel movement patterns for large brown trout during the summer. Clapp et al. (1990) suggested these behaviors reflected the rotation of foraging areas in response to a patchy environment. Although habitat selection is activity specific (Clapp et al. 1990), home locations typically selected by brown trout were relatively deep and slow moving with abundant cover (Heggenes 1988b; Clapp et al. 1990; Bunt et al. 1999). Large variability between or within the movement patterns of individual brown trout within a study was ubiquitous (Cresswell 1981; Clapp et al. 1990; Bunnell et al. 1998; Ovidio et al. 2002; Diana et al. 2004; Popoff and Neumann 2005; Heggenes et al. 2007). Authors have explicitly categorized movement behaviors of individual fish as short and long range displacement (Clapp et al. 1990), categorized movement strategies within populations as stationary or mobile (Diana et al. 2004), described a “two-step movement strategy” that included short-distance foraging activity and long displacements (Heggenes et al. 2007), or related movement strategies to brown trout grouped by survivorship (Bachman 1984; Aarestrup et al. 2005). Differences in movement behavior have also been attributed to variation in local habitat or food availability (Ovidio et al. 2002; Diana et al. 2004; Heggenes et al. 2007), to a shift in foraging strategy from drift feeding to piscivory (Clapp et al. 1990; Diana et al. 2004), or to fish size (Clapp et al. 1990; Bunnell et al. 1998; Heggenes et al. 2007) or origin (Bachman et al. 1984; Aarestrup et al. 2005). Physical characteristics of the local stream environment, such as temperature, flow and gradient, influence movement patterns of large brown trout. Reduced activity during high water temperatures or warm summer months has been observed (Cresswell 1981; Bachman 1984; Clapp et al. 1990; Popoff and Neumann 2005). Conversely, an increase in brown trout activity has been found with increasing river discharge (Clapp et al. 1990; Bunt et al. 1999; Popoff and Neumann 2005; Heggenes 62 et al. 2007). Diana et al. (2004) found that a stationary movement strategy was associated with steep gradient areas and a mobile strategy was associated with low gradient areas. In this study we characterize the dispersal, distribution, daily movements, and common habitats used by stocked large brown trout within three streams of the Upper Hudson River drainage – the Cedar, Indian and Hudson Rivers – during summer months. Methods We attempted to locate each fish during every day of tracking. The entire reach of each river was searched whenever possible, though weather conditions or logistical constraints infrequently prevented complete surveys. The collection of daily precise locations was attempted before the release time block so that differences between daily movement on release and non-release days could be assessed. Observations of fish locations were recorded automatically by an internal GPS within each portable ATS 4500S receiver, and therefore represent the location of the observer, not the fish. Observers moved along the river edge and marked the fish location at the point where the observer was standing on the bank perpendicular to the fish. The lowest gain setting that allowed for detection of the transmitter signal was recorded. This gain was then used to determine the accuracy range of the location observation. For the analysis, river bank locations were shifted to a corresponding midchannel location using GIS software. A river center line was digitized and segmented with a node at five meter increments, and then a nearest neighbor algorithm was used to shift each river bank location to the nearest node. Despite the measurement error and data manipulations, accuracy of measured locations was sufficient to place a fish within a geomorphic habitat unit. Data regarding the accuracy of the locations was not 63 consistently recorded during 2005, therefore the accuracy of these data was unknown and daily movement and dispersal were not estimated. Fish positions along the river center were logged as either meters upstream or downstream of the stocking location and used to calculate a number of spatial characteristics of each fish’s behavior. Dispersal, defined as the distance of individual fish from the stocking location at specified intervals following the stocking event (Bettinger and Bettoli 2002), was calculated approximately one day, one week, one month and two months after stocking in 2006. At each of these time intervals, the distributions of fish positions were designated by marking locations along each reach. Daily activity was calculated as the average distance traveled between daily locations (Bettinger and Bettoli 2002). Only observations on consecutive days were used in this calculation because a Spearman’s rank correlation revealed a bias in the movement data corresponding to the number of days between observations. A Kruskal-Wallis test (Hatcher and Stepanski 1994) was performed to determine if activity was different between rivers, and multiple comparisons were made using Dunn’s test (Zar 1996). The range of positions for an individual fish was calculated as the difference between the extreme upstream and downstream locations for 2006 (Appendix E) (Bettinger and Bettoli 2001). Results Initial dispersal and mean distance from the stocking location over time varied between the three reaches. Within the first week after stocking, brown trout in the Indian and Cedar Rivers dispersed immediately and were on average between 500 and 1000 meters from their respective stocking points (a deep pool in the Indian and a shallow run in the Cedar) (Table B.1). Within the Hudson River, trout remained relatively close to the stocking point (a cold water tributary) for the first day, but after 64 one week had dispersed over a much greater range than in either the Cedar or Indian Rivers. At longer time intervals after stocking (1 and 2 months), brown trout in the Cedar River remained similarly dispersed while those in the Indian were found closer to the stocking location than previously observed and those in the Hudson became less (after one month) and then more (after two months) dispersed. Table B.1. The mean ± 2SE (N) of the absolute distance (meters) from the stocking location for trout in the Cedar, Indian and Hudson River reaches at approximately one day, one week, one month and two months after stocking in 2006. Cedar Indian Hudson one day 934 ± 409 (11) 692 ± 610 (15) 302 ± 163 (15) one week one month 834 ± 409 (14) 1622 ± 681 (13) 815 ± 683 (14) 544 ± 445 (11) 1535 ± 1064 (12) 703 ± 622 (7) two months 1386 ± 716 (7) 80 ± 159 (2) 2681 ± 1984 (4) Similar to dispersal, the distribution of trout within the study reaches varied between rivers and over time. In the Cedar River, trout were evenly spread throughout the range of observed positions after initial dispersal and became more aggregated over time (Figure B.1a), with a single fish residing consistently 4 km upstream from the stocking site. Most trout in the Cedar River held positions that were at or downstream of where they were stocked, but in the Indian River nearly half of the trout seldom left the stocking location. Those that did move initially dispersed downstream and were positioned between 2 and 4 km from the stocking site (Figure B.1b). The location pattern after one week was very similar to one day post stocking, but at one month the trout were spread over a smaller range (~ 2.5 km) of positions, and some had moved upstream from the stocking location. All but four of the Hudson River trout dispersed from the stocking location within 48 hours, and movement was roughly equal in both the up and downstream directions (Figure B.1c). At one week, the trout had dispersed further upstream and downstream over a distance range of 65 Distance from stocking location (meters) Distance from stocking location (meters) 6000 4000 2000 0 -2000 -4000 -6000 1 day 1 week 1 month 6000 4000 2000 0 -2000 -4000 -6000 1 day 1 week 1 month 6000 4000 2000 0 -2000 -4000 -6000 1 day 1 week 1 month A. Cedar 2 months B. Indian 2 months C. Hudson 2 months Distance from stocking location (meters) Figure B.1a-c. Dispersal of brown trout in the Cedar (a), Indian (b) and Hudson (c) Rivers approximately one day, one week, and one and two months after stocking in 2006. The distance in meters up stream (positive) and down stream (negative) from the stocking location, is shown on the Y-axis. Each circle represents an individual fish. 66 approximately 10 km with four fish remaining close to the stocking point. At one month, locations of the remaining seven trout were spread over 3 km around the stocking location. After two months the four persisting trout were spread over a larger range of 6.5 km. The distribution of stocked trout was highly variable in the Hudson River and did not display a consistent pattern. The differences in distribution and dispersal of the trout in the three rivers can be explained, in part, by the daily movements of the fish (Figure B.2). Daily activity of trout in the Indian River was the lowest of the three rivers (median = 15 m) and the least variable. Activity of trout in the Cedar River was similarly low (median = 15m) with slightly greater variation, while trout in the Hudson River were considerably more active (median = 45 m). There was a significant difference in the daily activity of the trout between the three rivers (x2 = 29.00, p = <0.01). Activity of Hudson River fish was significantly different from both the Cedar and Indian Rivers (Q = 4.63 and 5.11, respectively at α = 0.05), and the activity of Indian River fish was not significantly different from those in the Cedar River (Q = 0.74). 10000 Daily distance moved (m) 1000 100 10 1 Cedar Indian Hudson Figure B.2. Activity level of brown trout in the Cedar, Indian and Hudson Rivers during the 2006 study period. Crosses represent the median, vertical lines extend to the maximum and minimum values, and grey boxes represent the middle 50% of the observations. The Y-axis is on a log scale. 67 The greater variability in activity observed in the Cedar and Hudson Rivers was driven by a few individual trout making sporadic long distance movements. In the Indian River, ranges in individual fish position less than 1 km were most common (10 fish) and the maximum observed range was just over 4 km (Appendix E). In the Cedar River, ranges in individual fish position greater than 10 km (2 fish), greater than 4 km (2 fish), and less than 3 km for the remaining fish were observed (Appendix E). In the Hudson River greater ranges in individual fish position were the norm. Ranges greater than 4 km for nine fish and less than 1 km for five fish were observed. Although the greatest ranges in position were observed in the Cedar River, trout in the Hudson River were observed traveling beyond our upstream monitoring site, but no precise locations were obtained. Brown trout frequently used specific locations in all three river reaches. In the Cedar River, all surviving fish after two months were in one of three locations: a deep pool just downstream of the Cedar River Dam (angling revealed that this pool held other stocked brown trout as well as brook trout), a run downstream of a cold tributary and with abundant overhanging vegetation, or in a groundwater fed plunge pool 4 km upstream from the initial stocking site (Figure B.3a). Before eight of the study trout passed downstream of a dam, observations of trout within another pool and a long, deep spring-fed run were common. In the Indian River most trout inhabited a few specific habitats that included the stocking pool, a tributary 2 km downstream from the stocking point, a deep pool-glide sequence below rapids 250 meters upstream of the stocking pool and a deep pool at the base of rapids 500 meters upstream from the stocking pool (Figure B.3b). An additional location (~1750 m in Figure B.3b) associated with a large debris jam was used frequently in 2005. This structure was removed late in the summer of 2005 and was therefore unavailable during 2006. Although there were locations regularly inhabited in the Hudson River, it was not 68 uncommon to observe trout outside of these habitats. Additionally, most Hudson River trout were observed alone, and when aggregations were observed they generally consisted only of pairs of trout. Locations associated with coldwater tributaries were most frequently used, especially within the coldwater plumes at the stocking tributary (Raquette Brook) and Aldous Brook. A deep pool known as the Black Hole and an area near groundwater seepage downstream of the Boreas River were also used regularly. Number of observations 0 250 Distance downstream of upstream-most observation (meters) 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Cedar River 200 150 100 50 0 Figure B.3a. Observed locations of brown trout in the Cedar River during 2006. The number of observations of study trout locations plotted in 300 m bins as a function of the distance downstream from the most upstream observed fish position. Tributaries are marked with a wavy line, pools with a circle, deep runs with an oval, the Cedar River Dam with a vertical bar and the stocking location with an arrow. 69 Number of observations Distance downstream of upstream-most observation (meters) 150 450 750 1050 1350 1650 1950 2250 2550 2850 3150 3450 3750 4050 4350 350 Indian River 300 250 200 150 100 50 0 Figure B.3b. Observed locations of brown trout in the Indian River during both 2005 and 2006. The number of observations of study trout locations plotted in 300 m bins as a function of the distance downstream from the most upstream observed fish position. Tributaries are marked with a wavy line, pools with a circle, and the stocking location with an arrow. 70 Number of observations 0 150 200 Distance downstream of upstream-most observation (meters) 1000 2000 3000 4000 5000 6000 7000 8000 9000 10000 Hudson River 100 75 50 25 0 Boreas River Black Hole Griffin Brook Raquette Brook 13 thAlLdaokuesOBurtloeotk Balm of Gilead Figure B.3c. Observed locations of brown trout in the Hudson River during both 2005 and 2006. The number of observations of study trout locations plotted in 300 m bins as a function of the distance downstream from the most upstream observed fish position. Tributaries are marked with a wavy line, pools with a circle, and the stocking location with an arrow. 71 Discussion The variation between movement patterns of the large brown trout in our three study reaches was likely due to fish taking advantage of differences in local environmental conditions, including habitat structure, food availability, temperature regime and gradient, and to individual variation in foraging behavior. Although we observed both mobile and stationary individuals (Diana et al. 2007) in all three study reaches, trout behavior in the Hudson, and to a lesser degree in the Cedar River, may be better described by two categories of movement for individual fish (Clapp et al. 1990; Heggenes et al. 2007). Most fish in these reaches remained stationary for periods of time with sporadic or, in many Hudson River fish, more regular long distance displacements. In the Indian River, few fish made any long-distance movements. Some of the individual variation in movement that we observed within reaches may be attributed to different diets. Brown trout that feed primarily upon invertebrate drift exhibit less long-range movements than piscivorous individuals, and a transition from drift feeding to piscivory has been observed in brown trout larger than 350 mm (Clapp et al. 1990; Bunnell et al. 1998). Most of the brown trout in our study were within this transitional size range (length >350 mm). Differences in the food available within each reach may explain some of the observed between-river differences in movement. Many trout in the Cedar and Indian River were positioned downstream from impoundments which may have provided abundant food sources, especially in the Indian River where densities of drifting macroinvertebrates were greater than in either of the other two rivers (Randy Fuller, Colgate University, unpublished data). Stream gradient may also have influenced the movement of brown trout in our study reaches. The more active behavior of Hudson River trout may be attributed to the low gradient of this river section. The energetic cost of traveling within steep 72 gradient systems has been suggested as a limit to long-range movements (Clapp et al. 1990; Diana et al. 2005). Living within the steepest gradient of the three reaches, the Indian River fish were also the most sedentary. Indian River study trout made fewer movements than those in the other reaches and were rarely observed outside the 4 km reach despite the lack of cool water patches and unobstructed access to deep gorge habitat in the Hudson River only a couple kilometers downstream. Additionally, differences in available habitat may have led to between-river variation in the movement of trout in our study. The slow and deep waters associated with cover preferred by adult brown trout (Heggenes 1988b; Clapp et al. 1990; Bunt et al. 1999) were more available in the Cedar and Indian Rivers than the Hudson River study reach. Additionally, when inundated with a flow release pulse, slow water habitat was greatly reduced in the Indian and Hudson Rivers (Baldigo et al. in prep). Deep pool and run habitats were frequently used in both the Indian and Cedar Rivers, whereas locations associated with tributaries were more common in the Hudson River reach. The density of tributaries was greatest for the Hudson River (1 ¼ per km), and one tributary and ¾ tributary per kilometer entered the Cedar and Indian River reaches, respectively. Trout in the Hudson may have moved more frequently in search of deep water habitat and seemed to make movements from one tributary confluence to another. Increased movement by brown trout has been observed during rain events (Clapp et al. 1990). Although some brown trout in our study did make large movements associated with a 2006 flooding event, many remained stationary and we did not find a consistent pattern. As in other investigations, we did not observe largescale movements by large brown trout associated with hydropeaking (Bunt et al. 1999; Heggenes et al. 2007), but similar to these authors, we did observe fish in areas of 73 relatively low velocity – near river banks or in tributaries – during pulsed discharge events. Given large within-river variability in movement and the lack of similarity between the two affected reaches, underlying differences in local environmental variables were probably more important than the occurrence of recreational discharge events in determining the movement patterns of adult brown trout in our study. Yet we did not record micro-scale movements that may have been energetically costly and may have been more likely influenced by hydropeaking. Adult brown trout living in these thermally stressful rivers, especially the Hudson River reach where slow, deep habitat was lacking, may not be able to fulfill metabolic needs – especially for hatchery origin trout that have been shown to be inferior at energy minimizing behavior (Bachman 1984; Aarestrup et al. 2005). 74 APPENDIX C QUALITATIVE ANALYSIS OF THE EFFECT OF PULSED DISCHARGE EVENTS ON THERMAL BEHAVIOR OF 2-YEAR-OLD BROWN TROUT IN THE UPPER HUDSON RIVER DRAINAGE We used a qualitative approach to describe the effects of pulsed discharge events on behavioral thermoregulation by stocked adult brown trout in addition to the comprehensive statistical analysis previously described (Chapter 1). For this evaluation we compiled data from sampling events for which data were available regarding trout body temperatures before and during the occurrence of a discharge pulse at a fish’s location (or during a similar time on a non-release day), using data from both mobile tracking (2005 and 2006) and the stationary receiver on the Hudson River (2006). A total of 31 unique fish were observed on 38 different days for a total of 108 such sampling events (i.e. data from some fish were available from more than one date). For each fish day, observations of fish body temperature from 1 ½ hours prior to and following the onset of the discharge pulse at the fish’s location were extracted from the dataset and plotted against time. For non-release days, the same three hour window was used with the reference “onset” time determined by averaging times from release days. The plots were visually assessed and categorized as displaying an increase, decrease or no change in fish body temperature, where a change was at least 0.5°C (Figure C.1). Percentages of sampling events within each release day category (release, non-release, and flooding – defined as daily discharge greater than or equal to 73.6 cms at the USGS gage station 01315500) were calculated (Figure C.2). Only four sets of release span data were available from fish within the Cedar River, and these were 75 left out of the analysis. No attempt was made to account for biases arising from repeated measures from an individual fish or from more frequent sampling at locations near the fixed receiver. With the exception of the 2006 flooding events, all sampling events occurred on days when ambient river temperature exceeded 20°C. Temperature (°C) 30 27 24 21 18 15 11:00 11:40 12:20 13:00 13:40 Eastern Standard Time 14:20 15:00 Figure C.1. Categorization scheme for series of fish body temperature observations 1 ½ hours before and after the arrival of the release pulse at a fish’s location. An increase (black circle), decrease (white circle) and no change (asterisks) in fish body temperature are depicted. The body temperature of brown trout in the Indian River remained unchanged on days when no release pulse occurred (non-release and flooding categories), but on release days an increase in body temperature was observed for 20% of the sampling events as the release pulse passed the location of a specific fish. In the Hudson River, both increases and decreases in trout body temperature were observed for 10-20% of the sampling events on days when no release pulse occurred, and no apparent difference was observed between the frequency of body temperature increases or decreases. Conversely, on release days a change in fish body temperature was 76 observed for 59% of the sampling events. Increases in body temperature were more frequently observed (38%) than decreases (22%) for sampling events on release days in the Hudson River reach. % Observations 100 80 60 40 20 0 10 100 80 60 40 20 0 2 15 2 = Non-release Day Indian River 82 Hudson River 7 13 12 = Release Day 5 4 22 6 = Flooding Days Figure C.2. Summary of 108 individual series of fish body temperature observations 1½ hours before and after the arrival of the release pulse at a fish’s location. The percentage of sampling events determined to exhibit either an increase (up arrow), decrease (down arrow) or no change (equals sign) in fish body temperature is shown for non-release days, release days and during flooding. The numbers of observations are shown in white lettering at the base of each bar. Observations for 2005 and 2006 are grouped. 77 These results suggest that recreational releases did alter the thermal behavior of the study trout. Increases in body temperature during releases were most likely due to thermal refuge dilution, but alternatively may have been due to a trout moving from a thermal refuge area into the main channel. Decreases in trout body temperature were likely the result of study fish moving upstream into tributaries. Regardless of the cause of these behaviors, increases in activity and / or body temperature likely inflicted acute or cumulative detrimental effects to the trout. 78 APPENDIX D ESTIMATION OF MISSING RIVER TEMPERATURE DATA Gaps in river temperature data, resulting from battery failure or late installation of loggers, were estimated using regression equations based on the complete records of the neighboring river temperature loggers. Correlations between data from loggers with incomplete data sets and loggers with complete data sets were made during the time periods when data were available for both. The resulting parameter estimates were used to calculate values for the missing data (predicted temperature). Daily temperature fluctuations were different for locations of loggers with complete records and locations of loggers with missing data. To account for this daily oscillation, we included the time of day (where “time” is measured in minutes 1-1440) as a third degree polynomial function and the interaction between time of day (time) and the temperature measured at the logger with complete data (temp) available. The difference in daily variability was also greater depending on the magnitude of the river warming and cooling (regulated by environmental or climatic influences such as light or discharge), therefore the maximum daily temperature (mxT) and relative water depth (stage) were also included in the regression model. Data were missing for the most downstream logger (HR05) in the Hudson River during a period of time in both 2005 and 2006 (Table D.1). Data were missing for the most downstream logger (CR04) in the Cedar River during 2006 (Table D.1). In all cases, regressions were based on data from the next logger upstream (HR04 and CR03, respectively). Investigation of residuals showed that the prediction errors were generally centered at zero and fell mostly within 1°C (Figures D.1-3). 79 Table D.1. Dates of data missing from river temperature loggers. Loggers with incomplete records (missing logger), the loggers with complete records (complete logger) used to interpolate missing data, the year and dates of missing data, and the dates when complete data existed for both loggers that were used to predict missing river temperatures (prediction dates). missing logger HR05 HR05 CR04 complete logger HR04 HR04 year 2005 2006 missing dates from to 08/10 08/10 08/14 08/20 06/29 08/13 CR03 2006 06/20 07/08 prediction dates from to 07/25 08/13 06/01 08/14 07/09 06/28 09/08 08/19 Table D.2. ANOVA table, fit statistics, effects tests and parameter estimates with standard errors for the multiple regression model for the most downstream river temperature logger in the Hudson River in 2005. Source Model Error Corrected Total DF 7 1816 1823 Sum of Squares 7487.52 793.10 8280.61 Mean Square 1069.65 0.437 F Value 2449.23 Pr > F <0.01 R2 CV Root MSE Mean of predicted temperature (°C) 0.90 2.78 0.66 23.81 Source intercept time temp mxT stage time*temp time*time time*time*time DF Type III SS 1 1 61.30 1 440.66 1 0.01 1 202.47 1 51.14 1 50.36 1 89.63 Mean Square 61.30 440.65 0.01 202.47 51.14 50.36 89.63 F Value 140.36 1008.99 0.02 463.61 117.09 115.30 205.23 Pr > F <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 Estimate -1.665527900 -0.010503335 0.970633465 0.004178775 0.957757450 0.000396529 0.000007435 -0.000000005 Std Err 0.83831778 0.00088654 0.03055703 0.02676703 0.04448143 0.00003664 0.00000069 0.00000000 80 300 200 frequency 100 0 -1.875 -0.875 0.125 1.125 2.125 observed - predicted Figure D.1. Frequency distribution of the residuals of the interpolated values for missing temperature observations for the most downstream Hudson River temperature logger in 2005. Table D.3. ANOVA table, fit statistics, effects tests and parameter estimates with standard errors for the multiple regression model for the most downstream river temperature logger in the Hudson River in 2006. Source Model Error Corrected Total DF 6 2680 2686 Sum of Squares 6542.15 463.21 7005.36 Mean Square 1090.36 0.17 F Value 6308.52 Pr > F <0.01 R2 CV Root MSE Mean of predicted temperature (°C) 0.93 1.92 0.42 21.60 Source intercept Mean DF Type III SS Square 1 time 1 4.95 4.95 temp 1 583.06 583.06 mxT 1 8.76 8.76 time*temp 1 3.37 3.37 time*time 1 0.87 0.87 time*time*time 1 5.45 5.45 F Value 28.61 3373.45 50.69 19.47 5.03 31.55 Pr > F <0.01 <0.01 <0.01 <0.01 0.03 <0.01 Estimate -2.293925673 -2.293925673 1.015318075 0.091670850 -0.000062490 0.000000840 -0.000000001 Std Err 0.26273293 0.26273293 0.01748095 0.01287510 0.00001416 0.00000037 0.00000000 81 frequency 600 400 200 0 -3.2 -2.4 -1.6 -0.8 0.0 0.8 1.6 observed - predicted 2.4 Figure D.2. Frequency distribution of the residuals of the interpolated values for missing temperature observations for the most downstream Hudson River temperature logger in 2006. . Table D.4. ANOVA table, fit statistics, effects tests and parameter estimates with standard errors for the multiple regression model for the most downstream river temperature logger in the Cedar River in 2006. Source Model Error Corrected Total DF 6 2230 2236 Sum of Squares 6472.51 911.61 7384.12 Mean Square 1078.75 0.41 F Value 2638.86 Pr > F <0.01 R2 CV Root MSE Mean of predicted temperature (°C) 0.88 2.91 0.64 21.96 Source intercept time temp mxT time*temp time*time time*time*time Mean DF Type III SS Square 1 3.64 1 667.88 1 19.25 1 12.34 1 12.47 1 40.46 3.63 667.88 19.25 12.34 12.47 40.46 F Value 8.90 1633.77 47.09 30.19 30.51 98.96 Pr > F <0.01 <0.01 <0.01 <0.01 <0.01 <0.01 Estimate -1.046343988 0.001848565 0.999310748 0.085087880 -0.000119812 -0.000003523 0.000000003 Std Err 0.45463767 0.00061968 0.02472319 0.01239917 0.00002180 0.00000064 0.00000000 82 frequency 600 400 200 0 -3.9 -2.7 -1.5 -0.3 0.9 2.1 3.3 observed - predicted Figure D.3. Frequency distribution of the residuals of the interpolated values for missing temperature observations for the most downstream Cedar River temperature logger in 2006. 83 APPENDIX E TABLE OF STATISTICS FOR INDIVIDUAL TROUT IMPLANTED WITH RADIO TRANSMITTERS Summary of individual stocked brown trout implanted with radio transmitters and tracked in the Cedar, Indian and Hudson Rivers during 2005 or 2006. The following descriptions are included: the year and river stocked and tracked; a randomly assigned unique fish identifier; the unique radio frequency of the implanted transmitter; the length and weight of the trout at the time of surgery; the number of days the trout was alive and within the study reach (persistence); whether the fish lived, died or disappeared (fate); the final resting location of the transmitter (final location); whether or not the transmitter was recovered (transmitter recovery); the mean of the body temperatures observed for that fish only during the afternoon time period; and for 2006 only, the mean daily distance moved in meters and the range of positions (upstream-most position minus downstream-most position) for each fish. 84 Appendix E. Table of statistics for individual trout implanted with radio transmitters. year river fish ID radio frequency (Mhz) length (mm) weight (g) persistence (days) fate end date final location transmitter recovery mean fish temperature (°C) ± SD 2005 Hudson 1 150025 377 624 13 dead 7-Aug-05 flood zone recovered 25.1 ± 1.3 2005 Hudson 2 150064 410 898 12 dead 6-Aug-05 flood zone recovered 24.7 ±1.5 2005 Hudson 3 150106 369 650 12 missing 6-Aug-05 missing missing 24.5 ± 1.7 2005 Hudson 4 150147 356 663 6 missing 31-Jul-05 missing missing 25.8 (N=1) 2005 Hudson 5 150183 392 698 24 alive unknown main channel not recovered 25.5 ± 1.6 2005 Hudson 6 150226 376 689 12 dead 6-Aug-05 in woods not recovered 24.8 ± 1.8 2005 Hudson 7 150263 381 716 12 missing 6-Aug-05 missing missing 24.8 ± 1.7 2005 Hudson 8 150303 377 647 4 missing 29-Jul-05 missing missing 26.1 (N=1) 2005 Hudson 9 150344 382 721 4 dead 29-Jul-05 in woods not recovered 26.1 (N=1) 2005 Hudson 10 150385 395 699 4 missing 29-Jul-05 missing missing 26.3 (N=1) 2005 Hudson 11 150426 388 650 16 dead 10-Aug-05 flood zone recovered 26.1 ± 1.9 2005 Hudson 12 150464 372 642 12 dead 6-Aug-05 main channel not recovered 24.2 ± 1.2 2005 Hudson 13 150503 342 567 3 dead 28-Jul-05 in woods recovered 25.4 (N=1) 2005 Hudson 14 150546 418 985 12 dead 6-Aug-05 flood zone recovered 25 ± 1.6 2005 Hudson 15 150585 382 673 22 dead 16-Aug-05 main channel recovered 25.9 ± 1.5 2005 Indian 16 150046 368 633 1 dead 26-Jul-05 in woods not recovered 21.8 (N=1) 2005 Indian 17 150086 373 645 7 dead 1-Aug-05 main channel recovered 24.4 ± 0.7 2005 Indian 18 150125 390 751 20 dead 14-Aug-05 flood zone recovered 25.2 ± 1.3 2005 Indian 19 150166 380 670 16 dead 10-Aug-05 main channel recovered 24.6 ± 1.4 2005 Indian 20 150204 368 696 24 dead 18-Aug-05 in woods not recovered 25.2 ± 1.1 2005 Indian 21 150244 371 637 7 dead 1-Aug-05 main channel not recovered 24.3 ± 0.7 2005 Indian 22 150283 358 589 20 dead 14-Aug-05 burrow not recovered 25.4 ± 1.0 2005 Indian 23 150324 389 766 7 dead 1-Aug-05 in woods not recovered 24.7 (N=1) 2005 Indian 24 150363 366 655 6 dead 31-Jul-05 in woods not recovered 24.6 (N=1) 2005 Indian 25 150405 378 678 17 dead 11-Aug-05 in woods not recovered 24.4 ± 0.4 2005 Indian 26 150446 352 550 18 dead 12-Aug-05 main channel recovered 25 ± 1.2 2005 Indian 27 150486 358 595 1 dead 26-Jul-05 in woods not recovered 21.9 (N=1) 2005 Indian 28 150525 370 575 16 dead 10-Aug-05 flood zone recovered 24.8 ± 1.1 2005 Indian 29 150565 426 1010 20 dead 14-Aug-05 main channel recovered 25.1 ± 1.2 2005 Indian 30 150604 390 781 1 dead 26-Jul-05 main channel recovered 24.8 (N=1) 85 Appendix E continued. Table of statistics for individual trout implanted with radio transmitters. Year 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 2006 river Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Hudson Indian Indian Indian Indian Indian Indian Indian fish ID 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 radio frequency (Mhz) 150514 150573 150663 150744 150822 150893 150914 150974 151404 151444 151634 151755 151833 151915 151974 151813 151853 150272 150553 150633 150724 150803 150873 150952 length (mm) 391 391 344 382 452 388 333 364 348 358 345 347 344 361 344 397 352 397 370 320 403 399 375 390 weight (g) 828 727 . 791 737 872 463 565 588 554 544 499 525 601 544 895 552 839 643 429 915 885 708 807 persistence (days) 23 62 7 27 67 20 15 20 35 37 7 67 26 20 20 38 8 56 62 52 29 39 67 28 fate missing dead missing dead alive missing dead dead dead dead missing alive dead dead missing dead dead dead dead dead dead dead alive dead end date 7-Jul-06 15-Aug-06 21-Jun-06 11-Jul-06 unknown 4-Jul-06 29-Jun-06 4-Jul-06 19-Jul-06 21-Jul-06 21-Jun-06 unknown 10-Jul-06 4-Jul-06 4-Jul-06 17-Aug-06 18-Jul-06 9-Aug-06 15-Aug-06 5-Aug-06 13-Jul-06 23-Jul-06 unknown 12-Jul-06 final location missing main channel missing main channel main channel missing main channel flood zone in woods main channel missing main channel main channel flood zone missing main channel flood zone main channel flood zone main channel main channel rookery main channel main channel transmitter recovery missing recovered missing recovered not recovered missing recovered not recovered recovered recovered missing not recovered recovered recovered missing recovered not recovered recovered recovered not recovered recovered not recovered not recovered recovered mean fish temperature (°C) ± SD 21.2 ± 2.1 22.6 ± 2.1 20.7 ± 3.4 20.1 ± 2.4 22.6 ± 2.2 21.8 ± 2.2 22.3 ± 1.9 21.2 ± 1.9 20.9 ± 1.8 21.0 ± 2.5 15.9 (N=1) 21.3 ± 1.7 21.3 ± 1.9 22.3 ± 1.9 21.3 ± 3.0 20.9 ± 1.6 21.7 ± 0.9 23.1 ± 1.5 23. 0 ± 1.6 21.8 ± 1.4 22.3 ± 1.0 22.5 ± 1.6 22.9 ± 1.6 21.9 ± 1.0 daily movement (m) 642 531 198 1251 456 1346 122 98 132 336 0 327 542 887 1559 496 12 48 72 224 316 232 25 45 range of positions (m) 3399 5261 594 5851 4148 4656 794 454 749 3715 0 5962 7536 3903 7795 4358 45 387 1356 3259 3672 3193 273 295 86 Appendix E continued. Table of statistics for individual trout implanted with radio transmitters. 87 year river fish ID radio frequency (Mhz) length (mm) weight (g) persistence (days) fate end date final location transmitter Recovery mean fish Temperature (°C) ± SD daily movement (m) range of positions (m) 2006 Indian 65 151213 430 1095 55 dead 8-Aug-06 flood zone recovered 23.3 ± 1.6 201 4154 2006 Indian 66 151292 349 609 19 dead 3-Jul-06 main channel recovered 21.8 ± 1.0 112 906 2006 Indian 67 151552 360 599 6 dead 20-Jun-06 main channel recovered 21.5 (N=1) 44 184 2006 Indian 68 151613 365 671 56 dead 9-Aug-06 rookery not recovered 22.9 ± 1.6 56 617 2006 Indian 69 151733 366 680 19 dead 3-Jul-06 main channel recovered 21.6 ± 1.1 27 68 2006 Indian 70 151813 399 817 3 dead 17-Jun-06 angled recovered N=0 0 0 2006 Indian 71 151874 393 823 33 missing 17-Jul-06 missing missing 22.0 ± 0.9 51 387 2006 Indian 72 151955 373 552 36 dead 20-Jul-06 rookery not recovered 22.3 ± 1.5 14 73 2006 Cedar 73 150532 379 733 48 dead 1-Aug-06 main channel recovered 19.0 ± 1.5 53 778 2006 Cedar 74 150613 338 535 67 dead 20-Aug-06* burrow not recovered 21.4 ± 2.2 296 4191 2006 Cedar 75 150703 305 506 67 alive unknown alive not recovered 22.3 ± 2.2 52 517 2006 Cedar 76 150784 365 640 46 dead 30-Jul-06 main channel recovered 20.8 ± 2.0 123 2074 2006 Cedar 77 150842 376 691 49 dead 2-Aug-06 main channel recovered 21.1 ± 2.0 293 2271 2006 Cedar 78 151032 386 841 67 alive unknown alive recovered 21.6 ± 2.2 1227 15452 2006 Cedar 79 151254 354 637 67 dead 20-Aug-06* main channel recovered 20.5 ± 1.5 20 192 2006 Cedar 80 151275 397 723 67 alive unknown alive recovered 21.8 ± 2.1 573 12554 2006 Cedar 81 151423 364 689 67 alive unknown alive recovered 21.5 ± 2.3 103 2203 2006 Cedar 82 151585 382 783 67 dead 20-Aug-06* main channel not recovered 22.4 ± 2.4 171 4654 2006 Cedar 83 151654 380 787 31 dead 15-Jul-06 in woods recovered 20.8 ± 1.6 105 2089 REFERENCES Aarestrup, K., Jepsen, N., Koed, A., and Pedersen, S. 2005. ‘Movement and mortality of stocked brown trout in a stream’, J. Fish. Biol., 66, 721-728. Allison, P.D. 1995. Survival Analysis Using the SAS System: A Practical Guide. SAS Institute Inc., Cary, NC. Bachman, R. A. 1984. ‘Foraging behavior of free-ranging wild and hatchery brown Trout in a stream’, Trans. Am. Fish. Soc., 113, 1-32. Bain, M. B., Finn, J. T., Booke, H. E. 1988. ‘Streamflow regulation and fish community structure’, Ecology, 69, 382-392. Baird, O. E. and Krueger, C. C. 2003. ‘Behavioral thermoregulation of brook and rainbow trout: comparison of summer habitat use in an Adirondack river, New York’, Trans. Am. Fish. Soc., 132, 1194-1206. Baird, O. E., Krueger, C. C., Josephson, D. C. 2006. ‘Growth, movement, and catch of brook, rainbow, and brown trout after stocking into a large, marginally suitable Adirondack river’, North Am. J. Fish. Manage., 26, 180-189. Baldigo, B. P., Mulvihill, C. I., Ernst, A. G., Boisvert, B. A. ‘Effects of recreational flow releases on natural resources of the Indian and Hudson Rivers in the Central Adirondack Mountains, New York’, Scientific Investigations Report 2008-xxxx, U.S. Geological Survey, Troy, NY. Report (in prep). Berman, C. H. and Quinn, T. P. 1991. ‘Behavioral thermoregulation and homing by spring Chinook salmon, Oncorhynchus tshawytscha (Walbaum), in the Yakima River’, J. Fish. Biol., 39, 301-312. Bettinger, J. M. and Bettoli, P. W. 2002. ‘Fate, dispersal, and persistence of recently stocked and resident rainbow trout into a Tennessee tailwater’, North Am. J. Fish. Manage., 22, 425-432. Bradford, M. J. 1997. ‘An experimental study of stranding of juvenile salmonids on gravel bars and in sidechannels during rapid flow decreases’, Regul. Rivers: Res. Manag., 13, 395-401. Bunnell, D. B., Isely, J. J., Burrel, K. H., and Van Lear, D. H. 1998. ‘Diel movements of brown trout in southern Appalachian streams’, Trans. Am. Fish. Soc., 127, 630-636. Bunnell, D. B. and Isely, J. J. 1999. ‘Influence of temperature on mortality and retention of simulated transmitters in rainbow trout’, North Am. J. Fish. Manage., 19, 152-154. Bunt, C. M., Cooke, S. J., Katopodis, C., and McKinley, R. S. 1999. ‘Movement and summer habitat of brown trout (Salmo trutta), below a pulsed discharge hydroelectric generating station’, Regul. Rivers: Res. Manag., 15, 395-403. 88 Burnham, K. P. and Anderson, D.R. 2002. Model Selection and Multimodel Inference: a practical information-theoretic approach. Springer-Verlag, New York. Burnham, K. P. and Anderson, D. R. 2004. ‘Multimodel inference: understanding AIC and BIC in model selection’, Sociol. Method. Res., 33, 261-304. Calles, O., Nyberg, L., and Greenberg, L. 2007. ‘Temporal and spatial variation in quality of hyporheic water in one unregulated and two regulated boreal rivers’, River. Res. Applic., 23, 829-842. Clapp, D. F., Clark, R. D., Diana, J. S. 1990. ‘Range, activity and habitat of large, free-ranging brown trout in a Michigan Stream’, Trans. Am. Fish. Soc., 199, 1022-1034. Cresswell, R. C. 1981. ‘Post-stocking movements and recapture of hatchery-reared trout released into flowing waters – a review’, J. Fish Biol., 18, 429-442. Dare, M. R., Hubert, W. A., and Gerow, K. G. 2002. ‘Changes in habitat availability and habitat use and movements by two trout species in response to declining discharge in a regulated river during winter’, North Am. J. Fish. Manage., 22, 917-928. Diana, J. S., Hudson, J. P., and Clark, R. D. Jr. 2004. ‘Movement patterns of large brown trout in the mainstream Au Sable River, Michigan’, Trans. Am. Fish. Soc., 133, 34-44. Ebersole, J. L., Liss, W. J., and Frissell, C. A. 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, J. L., Liss, W. J., and Frissell, C. A. 2003a. ‘Cold water patches in warm streams: physiochemical characteristics and the influence of shading’, J. Am. Water Resour. As., 39, 355-368. Ebersole, J. L., Liss, W. J., and Frissell, C. A. 2003b. ‘Thermal heterogeneity, stream channel morphology, and salmonid abundance in northeastern Oregon streams’, Can. J. Fish. Aquat. Sci., 60, 1266-1280. Elliot, J. M. 1994. Quantitative Ecology of the Brown Trout. Oxford University Press, Oxford. Elliot, J. M. and Elliot, J. A. 1995. ‘The effect of the rate of temperature increase on the critical thermal maximum for parr of Atlantic salmon and brown trout’, J. Fish. Biol., 47, 917-919. Elliot, J. M. 2000. ‘Pools as refugia for brown trout during two summer droughts: trout responses to thermal and oxygen stress’, J. Fish Biol., 56, 938-948. 89 Flodmark, L. E. W., Urke, H. A., Halleraker, J. H., Arnekleiv, J. V., Vøllestad, and L. A., Poléo, A. B. S. 2002. ‘Cortisol and glucose responses in juvenile brown trout subjected to a fluctuating flow regime in an artificial stream’, J. Fish. Biol., 60, 238-248. Flodmark, L. E. W., Vøllestad, L. A., and Forseth, T. 2004. ‘Performance of juvenile brown trout exposed to fluctuating water level and temperature’, J. Fish. Biol., 65, 460-470. Garrett, J. W. and Bennett, D. H. 1995. ‘Seasonal movement of adult brown trout relative to temperature in a coldwater resevoir’, North Am. J. Fish. Manage., 15, 480-487. Hatcher, L. and Stepanski, E.J. 1994. A Step-by-step approach to using the SAS System for univariate and multivariate statistics. SAS Institute Inc., Cary, NC. Heggenes, J. 1988a. ‘Effects of short-term flow fluctuations on displacement of, and habitat use by, brown trout in a small stream’, Trans. Am. Fish. Soc., 117, 336-344. Heggenes, J. 1988b. ‘Physical habitat selection by brown trout (Salmo trutta) in riverine systems', Nordic J. Freshw. Res., 64, 74-90. Heggenes, J. 1996. ‘Habitat selection by brown trout (Salmo trutta) and young Atlantic salmon (S. salar) in streams: static and dynamic hydraulic modelling’, Regul. Rivers: Res. Manag., 12, 155-169. Heggenes, J., Omholt, P. K., Kristiansen, J. R., Sageie, J., Ørkland, F., Dokk, J. G., and Beere, M. C. 2007. ‘Movements by wild brown trout in a boreal river: response to habitat and flow contrasts’, Fisheries Manag. Ecol., 14, 333-342. Jepsen, N., Mikkelsen, J. S., and Koed, A. 2008. ‘Effects of tag and suture type on survival and growth of brown trout with surgically implanted telemetry tags in the wild’, J. Fish. Biol., 14, 333-342. Jobling, M. 1981. ‘The influence of feeding on the metabolic rate of fishes: a short review', J. Fish. Biol., 18, 385-400. Jobling, M. 1997. ‘Temperature and growth: modulation of growth rate via temperature change’, in Global Warming - Implications for Freshwater and Marine Fish. Cambridge University Press, Cambridge. McDonald, D. G. & Wood, C. M. (Eds), pp. 225–253. Jepsen, N. Mikkelsen, J. S., and Koed, A. 2008. ‘Effects of tag and suture type on survival and growth of brown trout with surgically implanted telemetry tags in the wild’, J. Fish. Biol., 72, 594-602. Johnstone, H. C. and Rahel, F. J. 2003. ‘Assessing temperature tolerance of Bonneville cutthroat trout based on constant and cycling thermal regimes’, Trans. Am. Fish. Soc., 132, 92-99. 90 Kaeding, L. R. 1996. ‘Summer use of coolwater tributaries of a geothermally heated stream by rainbow and brown trout, Oncorhynchus mykiss and Salmo trutta’, Am. Midl. Nat., 135, 283-292. Lindstrom, J. W. and Hubert, W.A. 2004. ‘Mink predation on radio-tagged trout during winter in a low-gradient reach of a mountain stream, Wyoming’, West. N. Am. Naturalist., 64, 551-553. Littell, R.C., Milliken, G.A., Stoup, W.W., Wolfinger, R.D. 1996. SAS System for Mixed Models. SAS Institute, Inc., Cary, NC. Magnuson, J. J., Crowder, L. B., and Medvick, P. A. 1979. ‘Temperature as an ecological resource’, Am. Zool., 19, 331-343. Matthews, K. R., Berg, N. H., Azuma, D. L., and Lambert, T. R. 1994. ‘Cool water formation and trout habitat use in a deep pool in the Sierra Nevada, California’, Trans. Am. Fish. Soc., 123, 549-564. Moog, O. 1993. ‘Quantification of daily peak hydropower effects on aquatic fauna and management to minimize environmental impacts’, Regul. Rivers: Res. Manag., 8, 5-14. Nielsen, J. L., Lisle, T. E., and Ozaki, V. 1994. ‘Thermally stratified pools and their use by steelhead in northern California streams’, Trans. Am. Fish. Soc., 123, 613-626. Ovidio, M., Baras, E., Goffaux, D., Giroux, F., and Philippart, J. C. 2002. ‘Seasonal variations of activity pattern of brown trout (Salmo trutta) in a small stream, as determined by radio-telemetry’, Hydrobiologia, 470, 195-202. Pedersen, S. S., Dieperink, C., and Geertz-Hansen, P. 2003. ‘Fate of stocked trout Salmo trutta L. in Danish streams: survival and exploitation of stocked and wild trout by anglers,’ Ecohydrology & Hydrobiology, 3, 39-50. Pert, E. J. and Erman, D. C. 1994. ‘Habitat use by adult rainbow trout under moderate artificial fluctuations in flow’, Trans. Am. Fish. Soc., 123, 913-923. Poole, G. C. and Berman, C. H. 2001. ‘An ecological perspective on in-stream temperature: natural heat dynamics and mechanisms of human-caused thermal degradation’, Environ. Manage., 27, 787-802. Popoff, N. D. and Neumann, R. M. 2005. ‘Range and movement of resident holdover and hatchery brown trout tagged with radio transmitters in the Farmington River, Connecticut’, North Am. J. Fish. Manage., 25, 413-422. Power, G., Brown, R. S., and Imhof, J. G. 1999. ‘Groundwater and fish – insights from northern North America’, Hydrol. Process., 13, 401-422. Quigley, J. T. and Hinch, S. G. 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. 91 Reynolds, W. W. and Casterlin, M. E. 1979. ‘Thermoregulatory behavior of brown trout, Salmo trutta’, Hydrobiologia, 62, 79-80. Ross, M. J. 1982. ‘Shielded-needle technique for surgically implanting radiofrequency transmitters in fish’, Prog. Fish. Cult., 44, 41-43. Saltveit, S. J., Halleraker, J. H., Arnekleiv, J. V., and Harby, A. 2001. ‘Field experiments on stranding in juvenile Atlantic salmon (Salmo salar) and brown trout (Salmo trutta) during rapid flow decreases caused by hydropeaking,’ Regul. Rivers: Res. Manag., 17, 609-622. Scruton, D. A., Ollerhead, L. M. N., Clarke, K. D., Pennell, C., Alfredsen, K., Harby, A., and Kelley, D. 2003. ‘The behavioral response of juvenile Atlantic salmon (Salmo salar) and brook trout (Salvelinus fontinalis) to experimental hydropeaking on Newfoundland (Canada) river’, River. Res. Applic. 19, 577-587. Scruton, D. A., Pennell, C. J., Robertson, M. J., Ollerhead, L. M. N., Clarke, K. D., Alfredsen, K., Harby, A., and McKinley, R. S. 2005. ‘Seasonal response of juvenile Atlantic salmon to experimental hydropeaking power generation in Newfoundland, Canada’, North Am. J. Fish. Manage., 25, 964-974. Singer, J. D. 1998. ‘Using SAS PROC MIXED to fit multilevel models, hierarchical models, and individual growth models’, J. Educ. Behav. Stat., 24, 323-355. Skurdal, J., Hegge, O., and Hesthagen, T. 1989. ‘Exploitation rate, survival and movements of brown trout (Salmo trutta L.) stocked at Otta, southern Norway’, Regul. Rivers: Res. Manag., 3, 247-253. Summerfelt, R.C. and Smith, L.S. 1990. ‘Anesthesia, surgery and related techniques’ in Shcreck, C.B. and Moyle P.B (Ed.), Methods for Fish Biology. American Fisheries Society, Bethesda, MD. Pp. 213-272. Sutton, R. J., Deas, M. L., Tanaka, S. K., Soto, T., Corum, R.A. 2007. ‘Salmonid observations at a Klamath River thermal refuge under various hydrological and meteorological conditions’, River Res. Appl., 23, 775-785. Spigarelli, S. A., Thommes, M.M., Prepejchal, W., and Goldstein, R. M. 1983. ‘Selected temperatures and thermal experience of brown trout, Salmo trutta, in a steep thermal gradient in nature’, Env. Biol. Fish., 8, 137-149. Tate, K. W., Lancaster, D. L., and Lile, D. F. 2007. ‘Assessment of thermal stratification within stream pools as a mechanism to provide refugia for native trout in hot, arid rangelands’, Environ. Monit. Assess., 124, 289-300. Thomas, D. A. and Hayes, D. B. 2006. ‘A comparison of fish community composition of headwater and adventitious streams in a coldwater river system’, J. Freshwater Ecol., 21, 265-275. Torgersen, C. E., Price, D. M., Li, H. W., and McIntosh, B. A. 1999. ‘Multiscale thermal refugia and stream habitat associations of Chinook salmon in northeastern Oregon’, Ecol. Appl., 9, 301-319. 92 Valentin, S., Lauters, F., Sabaton, C., Breil, P., and Souchon, Y. 1996. ‘Modelling temporal variations of physical habitat for brown trout (Salmo trutta) in hydropeaking conditions’, Regul. Rivers: Res. Manag., 12, 317-330. Webb, B. W. and Walling, D. E. 1996. ‘Long-term variability in the thermal impact of river impoundment and regulation’, Appl. Geogr., 16, 211-227. Wehrly, K. E., Wang, L., and Mitro, M. 2007. ‘Field-based estimates of thermal tolerance limits for trout: incorporating exposure time and temperature fluctuation’, Trans. Am. Fish. Soc., 136, 365-374. Zar, J. H. 1996. Biostatistical analysis, 3rd edition. Prentice Hall, Englewood Cliffs, New Jersey. 93