The Role of Temperature and Flow on the Migration of Chinook Salmon (Oncorhynchus tshawytscha) Smolts Gregory E. Sykes B.Sc, University of British Columbia, 1999 THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE IN NATURAL RESOURCES AND ENVIRONMENTAL STUDIES (BIOLOGY) The University Of Northern British Columbia December 2007 © Gregory E. Sykes, 2007 1*1 Library and Archives Canada Bibliotheque et Archives Canada Published Heritage Branch Direction du Patrimoine de I'edition 395 Wellington Street Ottawa ON K1A0N4 Canada 395, rue Wellington Ottawa ON K1A0N4 Canada Your file Votre reference ISBN: 978-0-494-48818-8 Our file Notre reference ISBN: 978-0-494-48818-8 NOTICE: The author has granted a nonexclusive license allowing Library and Archives Canada to reproduce, publish, archive, preserve, conserve, communicate to the public by telecommunication or on the Internet, loan, distribute and sell theses worldwide, for commercial or noncommercial purposes, in microform, paper, electronic and/or any other formats. AVIS: L'auteur a accorde une licence non exclusive permettant a la Bibliotheque et Archives Canada de reproduire, publier, archiver, sauvegarder, conserver, transmettre au public par telecommunication ou par Plntemet, prefer, distribuer et vendre des theses partout dans le monde, a des fins commerciales ou autres, sur support microforme, papier, electronique et/ou autres formats. The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. L'auteur conserve la propriete du droit d'auteur et des droits moraux qui protege cette these. Ni la these ni des extraits substantiels de celle-ci ne doivent etre imprimes ou autrement reproduits sans son autorisation. In compliance with the Canadian Privacy Act some supporting forms may have been removed from this thesis. Conformement a la loi canadienne sur la protection de la vie privee, quelques formulaires secondaires ont ete enleves de cette these. While these forms may be included in the document page count, their removal does not represent any loss of content from the thesis. Bien que ces formulaires aient inclus dans la pagination, il n'y aura aucun contenu manquant. Canada ABSTRACT For salmonids, the smolting process includes substantial morphological, physiological and behavioural changes all of which must coincide to ensure the greatest chance of survival in the marine environment. I used historical data and a controlled laboratory experiment to investigate the role of both temperature and flow on the migration of Chinook salmon (Oncorhynchus tshawytscha) smolts. A combination of temperature experience (accumulated thermal units; ATU) and flow discharge were best able to describe the observed migration patterns. In addition, ATU consistently performed better than daily mean temperature suggesting that temperature experience plays a larger role in the migration process than a temperature threshold. In a laboratory experiment, fish in tanks with increasing temperature showed an earlier physiological and behavioural response than those in constant temperature tanks. Flow velocity had an effect on the pattern of migration but not on the onset of movement or on the physiological changes. Failure to consider both parameters when making management decisions or manipulating one independent of the other can have an impact on when fish begin migrating as well as on the migration pattern. TABLE OF CONTENTS GENERAL INTRODUCTION 1 CHAPTER 1 8 Abstract 9 Introduction 10 Methods Data Collection Model Development Model Selection Predictive Ability 13 13 16 20 22 Results Model Coefficients Model Validation 24 26 27 Discussion Migration Controls Parameter Interpretation 33 34 37 CHAPTER 2 44 Abstract 45 Introduction 46 Materials and Methods Fish Maintenance Experimental Set-up Water Quality Behavioural Analysis Behavioural Quantification Change-Point Analysis Physiological Sampling Problems Encountered Data Analysis Modeling: Physiology and Behaviour Interaction 48 48 50 51 52 54 56 56 58 59 60 Results Temperature 62 62 iv Flow Length, Weight, and Condition Factor Movement Physiology Modeling 64 65 66 68 73 Discussion 76 GENERAL SUMMARY 84 REFERENCES 92 v LIST OF FIGURES Figure 1. Gill Na ,K -ATPase activity from juvenile Chinook salmon captured in the upper Nechako River in 2004 by rotary screw traps during migration. Values are means ± SE. The initial value differed significantly from the subsequent values (1 way ANOVA; p<0.001) 4 Figure 2. Location of the Nechako River in central British Columbia, Canada. Rotary screw traps (RST) were located at Diamond Island while the Water Survey of Canada (WSC) data were collected at Cheslatta Falls. Scale for the map is 1cm equals 10km. 14 3 Figure 3. Average flow (m /sec; dashed line) and average temperature (°C; solid line) of the upper Nechako River for the periods 1992-2004 as recorded from the Water Survey of Canada station at Cheslatta Falls (Station ID# 08JA017). Average Chinook numbers are from rotary screw trap (RST) capture data from the same period (•). 15 Figure 4. Observed (•) and predicted (•) probability of downstream juvenile Chinook salmon {Oncorhynchus tshawytscha) migration from the Nechako River in central British Columbia, Canada for 1992 and 2004. Observed data were from rotary screw trap (RST) captures between April 4th and July 20th. Predicted results are based on a logistic regression model generated using RST capture data in combination with ATU and flow data from the Nechako from 1993 to 2003 28 Figure 5. Observed (•) and predicted (•) counts of juvenile Chinook salmon {Oncorhynchus tshawytscha) for 1992 and 2004 from the Nechako River in central British Columbia, Canada. Observed counts were from rotary screw traps (RST) captures from April 4th to July 20th. Predicted results were based on a ZINB model generated using RST capture data in combination with ATU, flow, and spawner data for the Nechako River collected from 1993-2003 29 Figure 6. Residual analysis (observed - predicted count) of migrating juvenile Chinook salmon {Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada, for 1992 (•) and 2004 (o). Observed data were based on rotary screw trap (RST) captures while predicted values were generated from a ZINB model developed using RST capture data along with ATU, flow, and spawner data for the Nechako collected from 1993-2003. A value of zero equals perfect prediction while negative values mean over-prediction and positive values mean under-prediction. 31 Figure 7. Total predicted (white) and observed (black) Chinook smolt counts for the Nechako River for 1992-2004. Observed data were based on rotary screw trap (RST) capture data while predicted counts were generated by a ZINB model using RST capture data along with data on ATU, flow, and number of spawners from the Nechako River for 1993-2003 32 VI Figure 8. Representation of the experimental tanks used to monitor Chinook salmon smolt movements. An oval barrier was used to direct fish through the two rectangular PITtag antennas which were attached to a datalogger. Directional spray bars were used to control water velocities within each tank and well water and surface water were mixed to achieve temperature manipulations 53 Figure 9. Daily mean temperature (°C) for 2005 (solid) and 2006 (dashed) for tanks 1 and 2 (increasing temperature) and tanks 3 and 4 (constant temperature) 63 Figure 10. Accumulated thermal units for 2005 (solid) and 2006 (dashed) for tanks 1 and 2 (increasing temperature) and tanks 3 and 4 (constant temperature) 64 Figure 11. Mean daily downstream (D/S) movements per fish for Chinook smolts (Oncorhynchus tshawytschwa) in four experimental tanks. Vertical lines identify points where a change-point analysis identified a significant change in migration pattern (>99% confidence; 10 000 bootstraps) 67 + + Figure 12. Average gill Na ,K -ATPase activity (umol ADP/mg protein/h) values (± SE) from samples collected from Chinook smolts (Oncorhynchus tshawytschwa) in 2005 and 2006 from four experimental tanks. Square and diamond data points represent yearly averages (2005 and 2006, respectively) for sampling events where a significant difference was detected between the two years 69 Figure 13. Average plasma Cortisol (ng/mi) values (± SE) from samples collected from Chinook smolts (Oncorhynchus tshawytschwa) in 2005 and 2006 from four experimental tanks. Square and diamond data points represent yearly averages (2005 and 2006, repectively) for sampling events where a significant difference was detected between the two years 72 vu LIST OF TABLES Table 1. Parameters used in the development of the binary and count predictive models for Chinook salmon (Oncorhynchus tshawytscha) downstream migration from the Nechako River in central British Columbia, Canada, from 1993-2003 18 Table 2. Candidate predictive Chinook salmon {Oncorhynchus tshawytscha) downstream migration models and associated rationale for selection for the Nechako River in central British Columbia, Canada, based on data collected from 1993-2003 20 Table 3. Summary of model selection statistics for the candidate binary and count models to predict the downstream migration of juvenile Chinook salmon {Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada 25 Table 4. Logistic regression output for ATU + ATU + Flow binary model to predict the downstream migration of juvenile Chinook salmon {Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada 26 2 Table 5. ZINB regression output for ATU + ATU + Flow + Spawner count model to predict the downstream migration of juvenile Chinook salmon {Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada 27 Table 6. Parameters used in the ITMC analysis of Chinook smolt downstream migration (daily proportion of total movements) during a controlled experiment 61 Table 7. Summary of model selection statistics for the candidate models to describe the downstream migration pattern of juvenile Chinook salmon {Oncorhynchus tshawytscha) in a controlled laboratory experiment 74 LIST OF APPENDICES Appendix 1. Water Quality Data 98 Appendix 2. 3-Factor Analysis of Variance (ANOVA) tables (2005 and 2006 data combined). Factors = sampling event (7 levels), temperature (2 levels), and flow (2 levels) 99 vin ACKNOWLEDGMENTS The author wishes to thank the Nechako Fisheries Conservation Program for providing access to data collected by the Program. This work was supported by a Natural Science and Engineering Research Council (NSERC) Discovery Grant to Dr. J Mark Shrimpton. The author was supported by an NSERC Industrial Postgraduate Scholarship (IPS) in association with Triton Environmental Consultants Ltd. Fish for use in the experiment were provided by the Doug Little Enhancement Facility at Penny, BC. The staff of the Quesnel River Research Center provided extensive assistance with construction of the experimental setup as well as fish maintenance during the experiment. Dr. J Mark Shrimpton provided assistance throughout the study and his support and encouragement are greatly appreciated. I would also like to thank the members of my supervisory committee, Dr. Chris Johnson, Dr. Stephen Rader and Dr. Tom Watson, who provided many helpful suggestions. Lastly, Danielle Sykes provided assistance during the construction of the experimental tanks and collection offish samples and provided moral support throughout the process. IX General Introduction 1 In preparation for seawater entry, juvenile salmon undergo significant morphological, physiological, and behavioural changes (Hoar 1988; McCormick et al. 1998) referred to as the parr-smolt transformation or smolting. Morphologically, smolting fish become more silvery (due to increased deposition of guanine and hypoxanthine in the skin) and become longer in relation to their weight. The physiological changes that occur during smolting include increases in several hormones such as growth hormone (GH), insulin-like growth factor 1 (IGF-1), Cortisol, and thyroxine (T4) (Hoar 1988; Iwata 1995; McCormick et al. 1998; Beckman et al. 2000). Considerable investigation has been directed at determining what role these hormones play in the onset and regulation of smolting and several comprehensive reviews of the physiology of smolting have been written (see McCormick et al. 1987; Hoar 1988; McCormick 1995 for reviews). Perhaps the most substantial physiological changes smolting fish undergo are those associated with the requirement for increased hypo-osmoregulatory ability for seawater survival. The transition from a hypoosmotic, freshwater environment where essential ions are lost with the elimination of excess water, to a hyper-osmotic, marine environment where water must be conserved and excess ions excreted, results in significant change to the basic function of many organs, including the gut, gills, and kidneys. In freshwater, the kidneys produce a significant amount of urine and ions are actively absorbed through the gills to compensate for the loss. However, in the marine environment, fluid absorption in the gut is greatly increased to conserve water and the kidneys produce less urine, but at a higher concentration. The kidney, however, is not the major route of ion excretion; instead the majority of sodium and chloride is excreted via 2 the gills. Within the gills, the most notable change in preparation for seawater entry is the proliferation of mitochondria-rich chloride cells and an increase in Na+,K+-ATPase activity, which is the primary enzyme associated with the excretion of Na + and CI" (McCormick 1995). The presence of both chloride cells and Na+,K+-ATPase is critical to maintaining ion regulation (McCormick et al. 1998). As a result of its critical role in hypo-osmoregulation and its association with the parr-smolt transformation, gill Na+,K+-ATPase activity is widely accepted as a measure of smolting and a means by which researchers can assess readiness of juvenile salmonids to enter seawater (McCormick 1993). Gill Na+,K+-ATPase data collected from migrating Chinook in the upper Nechako River in central British Columbia, Canada (Figure 1) show that changes in gill Na+,K+-ATPase activity occur even when the fish are still over 1,000 km and several weeks away from the marine environment, suggesting that gill Na+,K+-ATPase is a reliable smolting measure even in populations with long migration routes. Associated with the morphological and physiological changes that occur during the parr-smolt transformation is a behavioural response during which smolts initiate a downstream migration from their natal, freshwater streams and rivers to saltwater feeding habitats in the ocean. Physiological changes associated with the parr-smolt transformation, however, are reversible. If adverse environmental conditions are encountered or if significant delays in the migration occur, the seawater tolerance of the smolts will be lost (McCormick et al. 1999; Shrimpton et al. 2000). The limited period during which juvenile salmon display elevated seawater tolerance has led to the suggestion of a physiological 3 "smolt-window". The "smolt-window" is a relatively short period of time during the spring when juvenile salmon have a greater chance of survival when they enter the marine environment (McCormick et al. 1999; Sigholt et al. 1995). The timing of migration is therefore crucial to ensuring that fish reach the marine environment when physiological changes are at their peak. If SI I I 5 J I + 1 4 ra O) 0 Apr 28 May 03 May 08 May 13 May 18 May 23 May 28 Date Figure 1. Gill Na+,K+-ATPase activity from juvenile Chinook salmon captured in the upper Nechako River in 2004 by rotary screw traps during migration. Values are means ± SE. The initial value differed significantly from the subsequent values (1 way ANOVA; pO.001). 4 The series of changes that anadromous fish undergo during the parr-smolt transformation are complex and the role that environmental factors play in both the onset and timing of smolting has been extensively studied. The majority of research, however, has focused on the initiation of the physiological changes and in particular the influence of photoperiod and temperature on the process. Lengthening photoperiod in the spring is the primary cue for physiological smolting (Hoar 1988; Duston and Saunders 1990; Stefansson et al. 1991; McCormick et al. 1995). Studies have shown that an accelerated, lengthening photoperiod can induce a smolting response earlier than under normal conditions (Zaugg and Wagner 1973). Similarly, a constant photoperiod (i.e. no increase) can prevent smolting (Thorarensen et al. 1989; McCormick et al. 2002; Stefansson et al. 2007). Water temperature has also been found to have an impact on smolting, however, rather than directly initiating the process it has been found to be important in determining the rate of the physiological changes taking place (McCormick et al. 1997; McCormick et al. 2002; Shrimpton and McCormick 2003). In particular, studies have shown that while increased temperature results in the earlier development of smolt characteristics, extended exposure to higher temperatures will result in the loss of those same characteristics (Zaugg et al. 1972; McCormick et al. 1999). Studies have also shown that low temperatures will limit the physiological response to increased photoperiod (McCormick et al. 2000). Other factors such as changes in flow and turbidity have been less well studied. In addition, since the bulk of the work completed has focused on Atlantic salmon (Salmo salar) there is a need to study other salmonid species. 5 Understanding the role of these environmental variables on both the physiological and behavioural changes associated with the parr-smolt transformation is crucial from a management perspective. For example, it has been shown that dams delay the migration of anadromous fish as they are forced to traverse extensive slow moving reservoirs before passing the dam itself (Raymond 1979). These delays in the migration may result in the reversal of smolt characteristics as fish do not reach the ocean within the "smolt-window". In addition to these direct impacts on migration, managed systems often have flow and temperature regimes that differ from the "pre-development" stage. Since anadromous fish in these systems have evolved to respond to changes in environmental cues during the spring to time the physiological transformation with the onset of migration, it is unknown what potential impact manipulation of temperature and flow may have. It seems reasonable to assume that synchronous changes in environmental variables such as photoperiod, temperature, flow and turbidity will provide the strongest stimulus for smolting and migration and there is some evidence that an increasing photoperiod regime in synchrony with a rise in temperature strengthens the physiological changes associated with smolting (Muir et al. 1994). Yet, little is known regarding the impact on smolting if these physical changes are not synchronous or do not occur. To assess the impact of alterations in environmental cues on physiological and behavioural changes associated with the parr-smolt transformation, I examined the role of 6 two primary variables: temperature and flow. My work evaluated the manipulation of these two variables on the migration timing of Chinook salmon smolts using both observational and experimental techniques. The observational study (Chapter 1) involved the use of 13 years of historical Chinook migration data from a flow-controlled river to develop a statistical model that allowed me to assess the role of each parameter on migration. The experimental study (Chapter 2) used a controlled laboratory experiment to examine the role of temperature and flow on the behavioural and physiological changes observed in smolting Chinook independently, but also in combination. My thesis concludes with a comparison of the results from both studies as well as a discussion of the management implications of the results. 7 Chapter 1 The role of temperature and discharge on the migration timing of Chinook salmon {Oncorhynchus tshawytscha) smolts from a flow controlled river: the Nechako River, British Columbia Abstract I used an Information Theoretic Model Comparison (ITMC) analysis to investigate the role of daily mean temperature, temperature experience (accumulated thermal units; ATU), photoperiod and flow on the timing of downstream migration of Chinook salmon {Oncorhynchus tshawytscha) smolts from the Nechako River in central British Columbia, Canada. The study was based on rotary screw trap capture data collected by the Nechako Fisheries Conservation Program from 1992 - 2004. These data were used to develop both binary (migration event or not) and count (total number of migrants) models that predicted downstream migration of salmon. A total of 11 models for each of the two data distributions were compared using Akaike's Information Criterion (AIC). Both analyses identified a combination of temperature experience (ATU) and flow as being best able to describe the observed migration patterns. In addition, both analyses suggested that increasing ATU had a positive influence on migration, while increasing flow had a negative influence. Photoperiod was not included in the selected model for either analysis suggesting that despite its importance to the onset of smolting, it is less involved in the migration pattern. Lastly, ATU was found to have more influence on migration than daily mean temperature. Two years of independent data were used to test the predictive ability of each model. The count model was able to accurately predict the general trend in migration and in particular the end of migration period, but was not able to predict the daily fluctuations in movement. Alternatively, the binary model was able to predict whether or not fish would migrate on a given day with an accuracy of 93% and 99% for the two years tested. 9 Introduction The roles that environmental variables play in the behavioural aspects of smolting (i.e. downstream migration) have been less well studied than their role in controlling physiological changes associated with smolting. As outlined by McCormick et al. (1998), it is generally thought that migration is a combination of internal developmental changes (growth, condition factor), environmental priming factors (photoperiod and temperature), which cue physiological changes related to increased saltwater tolerance, and environmental releasing factors (increased temperature, flow and turbidity), which trigger the actual migration. While increased photoperiod is important to ensure physiological changes necessary for seawater residency occur, other environmental changes that occur synchronously in the spring such as warming water temperatures, increased flow, and increased turbidity are necessary to ensure the fish arrive at the marine environment at the appropriate time, thus ensuring the greatest chance of survival (Zaugg et al. 1985). In addition, Zaugg et al. (1985) suggested that migration is necessary to allow for a more thorough development of seawater tolerance. In a comparison of spring Chinook salmon from the Columbia River that were maintained in a hatchery during the normal migratory period with those that completed the natural migration, Zaugg et al. (1985) found that the fish that migrated had 2.5 times greater gill Na+, K+-ATPase activity, suggesting greater seawater tolerance. Despite its importance, little information exists on factors that control timing of migration. Factors such as photoperiod, temperature and flow, variation in species and population, adult escapement, rearing history, and size of juveniles, have all been linked 10 to salmonid migration to some extent (Roper and Scarnecchia 1999). In addition, differences in migration distances among populations of the same species can further compound the problem of understanding migration cues within a species. Early theories on the migration of smolts suggested that downstream movement was due to the limited swimming ability of juveniles coupled with increased spring flow resulting in involuntary downstream displacement (Thorpe and Morgan 1978). However, more recent experiments on swimming performance of salmon smolts have not shown any impairment in swimming ability making it unlikely that downstream movement is involuntary (Peake and McKinley 1998). While a few studies have suggested that increased flow in the spring may serve to stimulate migration in specific populations of Atlantic salmon, most also identified increasing temperature as a potential controlling factor (Hvidsen et al. 1995, McCormick et al. 1998, Antonsson and Gugjonsson 2002). In addition, photoperiod, which has been shown to have a significant effect on the physiological changes associated with smolting, may also play a role in determining the onset of migration. Given the number of environmental factors that are changing in the spring when migration occurs, it is difficult to say what role a particular variable or combination of variables has on the process. Similarly, it is unknown what effect there will be if certain environmental cues are delayed, accelerated or asynchronous. These issues are particularly relevant in flow controlled systems where manipulations of the hydrograph can result in substantial changes to the flow regime of the 11 system. From a management perspective, it is crucial to have an understanding of what effect, if any, the establishment of particular flow regimes can have on the smolting process. The objective of this study was to investigate the relationship between smolt migration and changes in environmental variables by correlating 13 years of Chinook smolt migration data with changes in five variables: water temperature (°C), accumulated thermal units (ATU), flow discharge (m3/sec), spawner numbers and photoperiod (hours of daylight). The data used in this analysis were collected between 1992 and 2004 from the Nechako River, located in central British Columbia, Canada. The river has been flow-controlled since 1952 following the construction of the Kenny Dam. I used a model selection analysis, referred to as Information Theoretic Model Comparison (ITMC), to complete the analysis. This technique was chosen because it allows for a comparison of multiple, competing hypotheses or models to explain a single phenomenon and is particularly useful in complex systems where multiple variables may be involved. For these types of systems, this method is considered more powerful and effective than standard null-hypothesis tests (NHT) (Anderson et al. 2000). The ITMC approach is based on an analysis of a set of biologically relevant models and one of the major criticisms of the technique is that often too many models are tested (Guthery et al. 2005). To avoid this problem, I decided to include only parameters that have been shown to be involved in smolt migration. Two types of models were included in the analysis based on different data distributions, each of which focused on a different aspect of smolt migration. The first was a binary analysis (presence/absence data; 12 logistic regression), which focused on whether or not migration occurred on a given day. The second was a count analysis (number of migrants; negative binomial regression), which focused on the numbers of fish migrating on a given day. While both types of models have been used to a certain extent to explain migration patterns of specific populations of fish (e.g., Nielsen et al. 2004; Paragamian and Kruse 2001), their use in the context of ITMC is novel from a fisheries standpoint. This analysis will provide a better understanding of factors that influence smolt migration in a flow regulated river and specifically how yearly differences in flow and temperature can affect migration patterns. Methods Data Collection The data used in the generation of the predictive Chinook migration models were collected from 1992 to 2004 as part of the juvenile outmigration study by the Nechako Fisheries Conservation Program (NFCP). Under that program, three rotary screw traps (RST) were fished in the Nechako River at Diamond Island, located 81 rkm (river kilometer) downstream of the Kenny Dam (Figure 2). The three traps were established at the same location each year with one located along the right margin on the river and two located in the mid-channel. The installation and removal dates of the traps varied from year to year depending on river conditions. In general, the traps were fished from the beginning of April through the end of July. 13 «K\ Kf£M BRITISH COLUMBIA Prince % % H> Geojge «^>W MY B- \ \ © H ^ '. .-^ • - i Vancouv "s-^_ Figure 2. Location of the Nechako River in central British Columbia, Canada. Rotary screw traps (RST) were located at Diamond Island while the Water Survey of Canada (WSC) data were collected at Cheslatta Falls. Scale for the map is 1cm equals 10km. Each trap was checked twice daily throughout the sampling period, at 8:00 am and 7:00 pm. Counts of the number offish of each species were recorded and juvenile and adult fish of the same species were differentiated, as were fry and smolts (in the case of Chinook). The study focused on the total number of Chinook smolts (size range 70-110 mm) captured daily from the three traps and the daily mean temperature and flow data. Data from 1993 to 14 2003 (n = 1125 records) were used in the construction of the model while data from 1992 and 2004 (n = 217 records) were kept separate for use in model validation. Average Chinook counts for 1992-2004 along with average flow and temperature for the same period are shown in Figure 3. o m o CM CM O -m *~^\ o o CM *—" ro i_ a. £ oco IT c o O .*: O O o o o c LO O ^ in March 1 1 1 1 April 1 May 1 June 1 July 1 August 1 Date Figure 3. Average flow (m3/sec; dashed line) and average temperature (°C; solid line) of the upper Nechako River for the periods 1992-2004 as recorded from the Water Survey of Canada station at Cheslatta Falls (Station ID# 08JA017). Average Chinook numbers are from rotary screw trap (RST) capture data from the same period (•). 15 Model Development Two types of models were developed: binary and count. The binary model was constructed using logistic regression with the data coded based on presence (1) or absence (0) of migrating Chinook. The count model was constructed using the number of migrating Chinook captured on a given day. Depending on the distribution of the count-based data, one can fit a Negative Binomial Regression Model (NBRM) or Poisson Regression Model (PRM); however, the latter is not generally appropriate due to overdispersion of the data (i.e., variance not equal to the mean; Long and Freese 2006). A likelihood ratio test was used to test for overdispersion in order to determine whether the PRM or NPRM was more appropriate. An additional consideration when choosing the appropriate count model is the frequency of zero counts. Both NBRM and PRM tend to under-predict the occurrence of zero counts relative to the observed data for datasets with a large number of zero counts (Lord et al., 2005; Long and Freese 2006). As a result, zero-inflated versions of these models (ZINB or ZIP) are generally recommended. In addition to predicting zeros that occur during the migration period, these types of models take into account situations where migration will never occur (for example due to time of year) and include that component to help account for the excess zeros in the dataset. The parameter(s) used for this Always Zero group is referred to as the "Inflate Term". A Vuong Test was used to compare the output of a NBRM to that of a ZINB and determine if a zero-inflated model was necessary (Vuong 1989). All statistics were completed using Stata (ver. 9.2, Statacorp, 2006). 16 Parameters Choosing from a small set of predictor variables considered ecologically plausible (Table 1), I developed a set of candidate models that were hypothesised to explain juvenile Chinook migratory behaviour. In addition to the photoperiod, temperature and flow terms, year-to-year variation in migration numbers can partially be explained by the numbers of spawners for that brood year and therefore spawner count was also included. Brood year for 1+ Chinook in the Nechako River is two years prior to the date of migration; therefore, a fish that migrates in 1994 was spawned in the fall of 1992. Lastly, since the migration pattern of Chinook smolts from the Nechako River was known to be non-linear, a quadratic equation was used to describe those variables that have a non-linear distribution during the migration period such as temperature and ATU. The quadratic equation consists of both a linear component, which captures the increase in migration early in the year, and a squared component which enables the model to better describe the decline in migration. All parameters with the exception of "Spawners" were tested to see if use of a quadratic was appropriate, however, only "Temperature2" and "ATU2" resulted in a decrease in AICC score. 17 Table 1. Parameters used in the development of the binary and count predictive models for Chinook salmon (Oncorhynchus tshawytscha) downstream migration from the Nechako River in central British Columbia, Canada, from 1993-2003. Parameter Temperature2 (°C)* Accumulated Thermal Unit2 (ATU)* Description Squared daily mean temperature. Squared cumulative daily mean temperature to describe temperature experience. 3 Flow (m /s) Total discharge at trap site. Photoperiod (hours) Period from sunrise to sunset. Spawners Number of spawners observed upstream of the trap site (2 years previous). * Where a squared term is included, the linear term is included as well. Both mean daily water temperature data and flow data were collected from the Water Survey of Canada (WSC) station located upstream of the trap site near Cheslatta Falls (Station ID# 08JA017). Although this station is located approximately 70 rkm upstream of the trap site, it is the most reliable and continuous dataset available for the upper Nechako River. This station also provided a record of river temperatures prior to trap installation, which was important for calculating ATUs. In addition, a comparison of the data collected at the trap site and at the WSC station for days when both were available did not show a significant difference in mean daily temperature (p = 0.1) or flow (p = 0.08). After reviewing the historical data from 1987-2004, I decided to calculate ATU's beginning on March 9th, as this was the date, on average, where the mean daily water temperature first reached 1°C for the year. Photoperiod data were calculated for the latitude and longitude of the nearest residential community to the trapping site (Fort Fraser, BC; 54'04" N 124'33" 18 W). The number of spawners was determined by aerial surveys completed by Fisheries and Oceans Canada (FOC) during the spawning period each year. Models A total of 11 models were developed (Table 2) all of which represented biologically plausible hypotheses to explain juvenile Chinook migration. Tolerance scores were used to assess collinearity among the parameters included in each model (Menard 2001). Due to collinearity between temperature and ATU, no models were tested that contained both of these parameters. Since the number of spawners will result in year-to-year variation in migrants that may not otherwise be identified, this parameter was included in all of the count models. The number of spawners would not be expected to influence whether or not a fish migrates, however, and was not included in the binary model. 19 Table 2. Candidate predictive Chinook salmon (Oncorhynchus tshawytscha) downstream migration models and associated rationale for selection for the Nechako River in central British Columbia, Canada, based on data collected from 1993-2003. Model Temperature2 1 i Flow Photoperiod ATU2 Temperature2 + Flow Temperature2 + Photoperiod Flow + Photoperiod ATU2 + Flow ATU2 + Photoperiod ATU2 + Flow + Photoperiod Temperature2 + Flow + Photoperiod Rationale Assess role of temperature on migration. Assess role of flow on migration. Assess role of photoperiod on migration. Assess role of temperature experience on migration. Assess role of a combination of water temperature and flow on migration. Assess role of a combination of water temperature and photoperiod on migration. Assess role of a combination of flow and photoperiod on migration. Assess role of temperature experience and flow on migration. Assess role of temperature experience and flow on migration. Assess role of combination of temperature experience, flow and photoperiod on migration. Assess role of a combination of water temperature, flow and photoperiod on migration. Note: All Count models also include the parameter "Spawners" f Where a squared term is included, the linear term is included as well. { Each of the Count models were calculated using the "ATU" parameter in the inflation portion of the model to describe the excess zeros, which were primarily associated with sampling later in summer (end of June and July) when the majority offish had migrated. Model Selection Using the ITMC technique, an Akaike's Information Criterion (AIC) value for each candidate model was calculated and model selection was based on the lowest AIC value. I used the small-sample bias correction form of AIC (AICC) in place of the standard AIC value. AICC has been shown to converge to the standard AIC value as sample size increases 20 and as a result can be used in all situations (Burnham and Anderson 2004). The formula to calculate AICC is: AICC = -ILL + 2K + 2K( K +l ) n-K-\ where: LL = log likelihood, K = # of parameters, and n = sample size. Since the actual value of AIC (or AICC) for a particular model is less important than the change in AIC between models (Burnham and Anderson 2004) the term A, was calculated. This value represents the change in AIQ value between each model and the model with the lowest AICC score. The "best" model will therefore have a A, value of "0" whereas the other models will have positive values. This transformation represents the information lost if model, were used in place of models (Burnham and Anderson 2004). A general rule of thumb is that if A; < 2, the models are too similar to be ranked by AIC value alone and in such situations the most parsimonious one should be selected (Anderson et al. 2000). For the best binary and count models, a (3-coefficient was generated for each parameter, the sign of which corresponded to the direction of the effect. A z-statistic was also calculated to assess the significance of the individual parameters. Significance was determined as p < 0.05. 21 Predictive Ability A major criticism of ITMC methods is that often the "best" model is not validated using independent datasets (Guthery et al. 2005). As a result, the ability of the model to perform in real-world applications is unknown. To that end, Chinook migration data from the Nechako River from 1992 and 2004 were kept separate from those years used to generate the models (1993 - 2003). The withheld data were used to assess the predictive ability of the final models. For the binary analysis, I used a binary logit model to generate the predicted probability of Chinook migration: Y _ exp(/?0 + fl*i + P2x2 +... + /?,*,) 1 + exp(/?0 + /?,*, + /32x2 +... + p,x,) where, fy is the intercept term, /?,• is the coefficient for each covariate in the model, and xt is the value of the covariate. Pearson's standardized residuals were used to assess the difference between the observed and predicted values. Standardized residuals have a normal distribution and therefore should have a mean of 0 and a standard deviation of 1. In addition, 95% of the residuals should fall between -2 and 2 with larger and smaller values identifying cases where the model works poorly or that exert more than their share of influence on the model parameters (Menard 2001). A common means of assessing the accuracy of the predicted probabilities is to round the calculated value to either 0 or 1 based on a probability threshold, but model performance will vary depending on the threshold used and this technique is not recommended (Boyce et 22 al. 2002). Instead, a Receiver Operating Characteristic (ROC) curve was generated, which evaluates the proportion of correctly and incorrectly classified predictions over a continuous range of probability threshold levels from 0 to 1 (Pearce and Ferrier 2000). The area under the curve (AUC) provides an estimate of the overall predictive ability; a model with an AUC of 1.0 is a perfect predictor whereas a model that has no predictive ability has an AUC of 0.5 (Boyce et al. 2002). The general guidelines for interpreting the value of the AUC of a ROC curve in regards to predictive ability are: poor (0.5 - 0.7), reasonable (0.7 - 0.9), and very good (0.9 - 1.0) (Swets 1988). For the count analysis, I used the Stata program "prcounts.ado" (Long and Freese 2006) to predict the number of migrating salmon. Due to non-normality of the data, a nonparametric Wilcoxon rank-sum test was used to compare the observed and predicted counts for 1992 and 2004. Residuals, the difference between the observed and predicted counts, were also calculated. A mean of zero suggested perfect prediction whereas negative values suggested over-prediction (predicted value > observed value) and positive values suggested under-prediction (predicted value < observed value). 23 Results The negative binomial regression model (NBRM) was a preferable model to the Poisson regression model (PRM) because of overdispersed count data (G2=3180.01, p<0.001). In addition, when the NBRM and the ZINB were compared, the large number of zero counts resulted in a slightly better fit for the ZINB (Vuong = 0.98, p=0.16). Although not a significant difference, this result was confirmed by lower AIC scores and better predictive ability of the ZINB compared to the same model calculated using the NBRM. Based on AICC score, a combination of ATU2, flow and photoperiod was best able to model the probability of migration, while ATU2, flow and spawner numbers were best able to model the number of migrants (Table 3). For the binary analysis, the model ranked 2nd (ATU2 + flow) had an AICC score only 1.6 points greater than the model ranked 1st (ATU2 + flow + photoperiod), suggesting they are nearly equivalent. Since a model with one less parameter is likely a more parsimonious explanation of the observed data, I selected the model ranked 2nd (ATU2 + flow). Those same models (with the addition of the Spawner parameter) were also the top two for the count analysis with a difference in AIQ score of only 1.1 points. However, for that analysis the simpler version (ATU2 + flow + spawner) was ranked 1st, and was therefore selected. 24 Table 3. Summary of model selection statistics for the candidate binary and count models to predict the downstream migration of juvenile Chinook salmon (Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada. Binary Model Rank AIQ ATU + Flow + Photoperiod ATU2 + Flow ATU2 ATU2 + Photoperiod Temp2 + Flow Temp + Flow + Photoperiod Temp2 Temp2 + Photoperiod Flow + Photoperiod Photoperiod Flow Zero-Inflated Neg. Binomial Count Model 1 2 3 4 5 6 7 8 9 10 11 Rank 596.710 598.354 646.186 648.188 718.554 720.070 793.032 793.967 1098.142 1136.959 1376.201 AICC &AICc ATU + Flow + Spawner ATU2 + Flow + Photoperiod + Spawner Temp2 + Flow + Spawner Temp2 + Flow + Photoperiod + Spawner ATU2 + Photoperiod + Spawner ATU2 + Spawner Temp2 + Photoperiod + Spawner Flow + Spawner Temp2 + Spawner Flow + Photoperiod + Spawner Photoperiod + Spawner 1 2 3 4 5 6 7 8 9 10 11 4736.510 4737.637 4784.116 4785.487 4810.104 4810.571 4850.677 4852.214 4852.479 4854.113 4896.938 0.0 1.1 47.6 49.0 73.6 74.1 114.2 115.7 116.0 117.6 160.4 2 AAICC 0.0 1.6 49.5 51.5 121.8 123.4 196.3 197.3 501.4 540.2 779.5 nQ Beyond the models ranked 1st and 2-.nd , the rankings of the remaining models differ for the two distributions. In the case of the binary model, the parameter "ATU " was found in each of the top 4 models, with "temp2" found in the following 4 (ranked 5-8). In the case of the count model, the parameter "flow" was included in each of the top 4 models and similar 25 to the binary models, ATU2 was found to result in lower AIC scores than the corresponding temp model. Model Coefficients The coefficients generated for each of the parameters included in the selected binary (Table 4) and count (Table 5) models were all found to have a significant effect on migration (p<0.001). Pearson's standardized residuals for the binary analysis indicated that no records influenced the model disproportionately (mean = 0.002, standard deviation = 0.97, 96% of the values being between -2 and 2). For both the binary and count models, flow and the squared component of the ATU quadratic were found to have a negative influence on migration, while the linear component of the ATU quadratic, and spawner numbers (count model only) had a positive influence. Table 4. Logistic regression output for ATU + ATU2 + Flow binary model to predict the downstream migration of juvenile Chinook salmon (Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada. Parameter P Standard Error Z P ATU ATU2 Flow Constant 0.0117 -2 x 10"5 -0.0154 2.5336 0.0029 3.47xl0"6 0.0026 0.3571 4.70 -7.53 -5.98 7.10 <0.001 O.001 <0.001 <0.001 26 3 95% CI Lower Upper 0.0068 0.01656 -3.29xl0~5 -1.93xl0"5 -0.0204 -0.0103 1.8338 3.2334 Table 5. ZINB regression output for ATU + ATU2 + Flow + Spawner count model to predict the downstream migration of juvenile Chinook salmon (Oncorhynchus tshawytscha) from the Nechako River in central British Columbia, Canada. Count Portion ATU ATU2 Flow Spawner Constant Inttate Portion ATU Constant p 0.0083 -1.7xl0"5 -0.0119 0.0003 1.3985 P 0.0178 -11.1827 Standard Error 0.001 1.8xl0"6 0.0014 3.9xl0"5 0.1457 Standard Error 0.0032 1.8484 Z P 8.08 -9.31 -8.60 9.81 9.60 Z statistic O.001 O.001 <0.001 <0.001 O.001 p value 5.59 -6.05 O.001 O.001 p95%CI Lower Upper 0.0063 0.0103 -2.1xl0"5 -1.4x10 -0.0147 -0.0092 0.0003 0.0005 1.1129 1.6841 p 95% CI Lower Upper 0.0119 0.0241 -14.8056 -7.5599 Model Validation For both 1992 and 2004, independent migration data were an excellent fit to the best binary model (Figure 4). AUC was 0.93 and 0.99 for 1992 and 2004, respectively. For both years, the model accurately predicted the transition from continuous migration in early spring to no migration by mid June. 27 Iro Sco. 1. April 1 May 1 June 1 1992 July 1 August 1 April 1 May 1 June 1 2004 July 1 August 1 ih Sep O !o 0! 2^ Q. Figure 4. Observed (•) and predicted (•) probability of downstream juvenile Chinook salmon {Oncorhynchus tshawytscha) migration from the Nechako River in central British Columbia, Canada for 1992 and 2004. Observed data were from rotary screw trap (RST) captures between April 4th and July 20th. Predicted results are based on a logistic regression model generated using RST capture data in combination with ATU and flow data from the Nechako from 1993 to 2003. Results of the model validation for the count model (Figure 5) show that for 1992 the model tended to over-predict the migration numbers, while for 2004, the model results were 28 more representative of mean migrant numbers. For both years, the model was not able to recreate the daily fluctuations in migrant numbers, but was successful at identifying the end of the migration period. 50. O 0.5 m/s) scenarios. Combinations of the manipulations were established in four experimental tanks. At two-week intervals tissue and blood samples were collected from a random group of fish from each tank for analysis of physiological changes associated with smolting (gill Na+,K+-ATPase activity and plasma Cortisol concentration). Fish from the tanks with increasing temperature showed an earlier peak in movement than those from the colder tanks and also showed a peak in gill Na+,K+-ATPase activity, as opposed to a steady increase for the duration of the experiment. Flow velocity did not influence physiological parameters associated with smolting and was not found to initiate movement as was the case with accumulated thermal units (ATU). However, the presence of a strong, directional flow did result in a period of more defined movement suggesting a possible influence once migration is underway. A model was used to correlate the observed fish movements with environmental parameters and it was found that a combination of either photoperiod or ATU with gill Na+,K+-ATPase activity were most strongly linked to 45 movement. ATU was also found to be more strongly correlated with the smoking process than was daily mean temperature. Introduction It is generally accepted that deviations in either the timing or duration of the migration of salmonid smolts can have serious implications on potential survival in the marine environment. The period when the physiological changes associated with smolting are at their peak has been shown to be relatively short and can be influenced by environmental conditions such as temperature and photoperiod (McCormick et al. 1999). As a result, migrations that begin too late or that take longer than normal can result in fish arriving at the marine environment when physiological changes are not at their peak. An understanding of how manipulation of environmental variables can impact the onset, rate and duration of migration is therefore important from a management perspective. This is particularly true in flow-controlled systems where flow regimes that might maximize hydroelectric production are substantially different and often inverted from "normal" conditions, to which fish have naturally evolved. Strategies to manage the systems such that conditions do not adversely affect fish are further complicated by confounding flow requirements of different species and life stages which also might require management. Several studies on wild populations of juvenile salmonids have shown strong correlations between migration timing and both increasing water temperature (e.g. Jonsson 46 1991; Hvidsten et al. 1995; Roper and Scarnecchia 1998; Whalen et al. 1999), and increasing flow (Hvidsten et al. 1995; Baker and Morhardt 2001). In addition, increases in both temperature and flow have been shown to increase the rate of migration in juvenile Chinook salmon (Raymond 1979; Conner et al. 2003) suggesting a possible explanation why populations may have adapted to use these factors as releasing stimuli for migration. Raymond (1979) found that in the Snake River, years with lower flow resulted in slower rates of migration, delayed migration and, ultimately, cessation of movement of some individuals. Similarly, it has been shown that dams prolong migration as fish are forced to traverse extensive slow moving reservoirs before passing the dam itself (Raymond 1979). However, while the behavioural side of smolting (migration) may be delayed, the physiological side may still proceed at a normal rate resulting in a less than optimal situation where physiological changes have begun to revert upon arrival at the marine environment. As a result of the potential impact of delayed migration, flow augmentation is a management strategy recommended by many researchers (Connor et al. 2003; Smith et al. 2003) as a means of increasing migration rate. However, failure to consider the timing of the flow increase or the effect of the increase on other parameters such as temperature, may limit the benefits. Despite the work that has been completed on understanding the role of environmental parameters such as flow and temperature on smolt migration, few studies have involved direct manipulation of these conditions in a controlled laboratory experiment. Without the 47 ability to control variables while manipulating others, it is difficult to assess their relative importance to the migration process. For example, Zydlewski et al. (2005) experimentally manipulated temperature and found that cumulative temperature was more of a factor in initiating and terminating migration in Atlantic salmon than was a particular temperature threshold. However, other factors, such as increased flow, were not included in that study. Given the potential importance of flow to the migration process and the fact that flow augmentation is a management strategy recommended by some researchers, there would appear to be a need for additional experimental research in this area. This study used a controlled laboratory experiment to investigate how manipulations of flow velocity and temperature influenced the downstream movements associated with smolt transformation of juvenile Chinook salmon. The results provide insight as to how manipulations of those variables might influence the behavioural and physiological changes associated with the smolting process. Materials and Methods Fish Maintenance Chinook salmon parr ranging in size from 80 - 110 mm were transported from the Doug Little Enhancement Facility, Penny, BC (Dome Creek stock) by truck to the Quesnel River Research Center (QRRC) in Likely, BC in 2005 (March 15th; n=325) and 2006 (March 7th; n=315). Similar to the Nechako River, Dome Creek is an upper Fraser River drainage however, it is located approximately 210 km upstream of the Nechako confluence. Aerators 48 were used throughout the 4-hour transport to ensure a continuous supply of oxygen. Dissolved oxygen (DO) and temperature were monitored and maintained at levels approximating that of holding conditions at the Penny Hatchery (DO = 12 mg/L; temperature = 1°C). No mortalities occurred during transport in either year. At the QRRC, fish were transferred into a single 1.8 m diameter circular tank with a continuous supply of river water from the Quesnel River (DO = 12mg/L; temperature 2.5°C). Photoperiod was maintained at natural daylength for the latitude of Dome Creek (53° 5') using a programmable timer on overhead incandescent lights. Fish were manually fed pellet food (EWOS 1.5, Vancouver, BC) daily to satiation. Approximately 2 weeks after transport, fish were tagged internally with a 23 mm long, 3.4 mm diameter, 0.6 g passive integrated transponder (PIT) tag with a unique 8-digit code (Texas Instruments, Piano, TX). Prior to tagging, fish were anaesthetized with 100 mg tricane methane sulfonate (MS-222)/L buffered with sodium bicarbonate (pH = 7.0) and the fork length (mm) and weight (g) of each was determined. A small incision (approximately 4 mm) was made in the ventral surface of the fish between the pectoral and pelvic fins and the tag was inserted. Vetbond™ (3M, London, ON) surgical adhesive was used to close the wound. Fish were allowed to recover fully from the procedure and were then randomly and evenly distributed into each of the four experimental tanks (n = 75/tank). 49 Experimental Set-up Circular fibreglass tanks (green, 1.8 m diameter, 0.9 m deep, and volume of approximately 2.4 m ) were used to allow for the establishment of directional flow and each was plumbed to provide surface water (Quesnel River; natural temperature regime) and/or well water (constant 6°C). The well was located at the QRRC and was previously used to supply water when the site was a functioning hatchery. Temperature manipulations were achieved by mixing surface water and well water. Spray bars constructed from PVC were used to establish a directional flow within each tank and velocity was manipulated by changing the angle of the spray-bar while maintaining a constant volume of water entering each tank. A PVC standpipe drain in the center of each tank (within the oval of the barrier to prevent fish entrainment) allowed for a flow-through set-up. For one week following tagging, each of the 4 tanks was maintained on exclusively river water and low velocity discharge to ensure complete recovery from the tagging procedure. After the first week, the following experimental regimes were established by manipulating temperature and flow within each of the tanks: 1. Increasing temperature regime; low velocity flow; 2. Increasing temperature regime; high velocity flow; 3. Constant temperature, high velocity flow; and 50 4. Constant temperature with low velocity flow. All four tanks experienced the same naturally increasing photoperiod and were manually cleaned at weekly intervals using siphon tubes, dip nets and brushes. Water Quality A comparison of general water quality parameters between the well and river water was completed in the spring of 2005 to determine what influence, if any, differences might have on the behavioural and physiological responses. Duplicate 250 ml samples were collected from each source and were sent to the University of Victoria, School of Earth and Ocean Sciences Laboratory for analysis. Conductivity and pH measurements were collected using a hand-held meter (Hanna Combo, Hanna Instruments, Laval, Quebec). Water quality samples were analyzed for a range of common elements and no substantial differences were identified that would be expected to affect physiological or behavioural changes associated with smolting (see Appendix 1). Measurements of pH and conductivity (uS) were collected during each sampling event and were found to be similar between the two sources (mean pH: river 7.1 + 0.03, well 7.0 + 0.01; mean conductivity: river 96 + 7.8 uS, well 114 + 1.2 uS). Water temperatures were measured hourly using Hobo water temperature loggers (Onset Computer Corporation, Bourne, MA). 51 Behavioural Analysis Following the procedure outlined by Zydlewski et al. (2005) each of the experimental tanks had an oval barrier installed in the middle to force fish to travel around the outside of the tank (Figure 8). At the narrowest point, the distance between the barrier and the edge of the tank was 0.45 m while at the widest was 0.9 m. To monitor movements, two rectangular antennas (0.45 m x 0.60 m) were built to fit into the narrow portion of each tank, angling toward the center of the tank. The antennas were positioned to be at least 0.50 m apart and were connected to separate tuners. Each tuner was in turn connected to a multiplex reader (Oregon RFID, Portland, Oregon) with a PDA datalogger. Multiplex readers could read 4 antennas and therefore one unit was used for tanks 1 & 2 and a second for tanks 3 & 4. When in the field of an antenna, the individual tags are energized and the antenna number, time, date, and tag number are logged. The direction of travel of a specific fish could therefore be determined by looking at the order in which an individual tag was read by subsequent antennas within a set time period. 52 Well Surface Water Water Rectangular Antennas 1/ Multiplex Datalogger I Figure 8. Representation of the experimental tanks used to monitor Chinook salmon smolt movements. An oval barrier was used to direct fish through the two rectangular PIT-tag antennas which were attached to a datalogger. Directional spray bars were used to control water velocities within each tank and well water and surface water were mixed to achieve temperature manipulations. Tanks with increasing temperature (1 & 2) were on exclusive river water (approx. 2°C) for one week post-tagging (mid/end of March) and were then gradually switched to exclusive well water (6°C) by the first week of April. The tanks were maintained at this temperature until the river water reached 6°C (beginning of May) at which time they were 53 switched back to exclusive river water and allowed to increase steadily for the remainder of the experiment. The result was a temperature regime approximately 4°C warmer than river conditions for the month of April followed by a steady increase in temperature for the remainder of the experimental period. The tanks with constant temperature (3 & 4) were maintained on exclusively river water until the temperature was approximately 6°C at which point they were switched to exclusively well water for the remainder of the experiment. The result was a temperature regime that increased naturally to 6°C and was then held steady at that level. Flow velocity within all 4 tanks was initially maintained at between 0.05 -0.1 m/s (referred to as "low velocity") for 2 weeks following the tagging, while the fish recovered from the procedure. Following recovery, the flow velocity in tanks 2 and 3 were increased to 0.5 m/s (referred to as "high velocity") while tanks 1 and 4 were maintained at low velocity, which was considered sufficient to provide directional flow in the tank only. Throughout the experiment, netting was used to cover each of the four tanks with black plastic around the periphery to provide shading. No substrate was placed in the tanks. Behavioural Quantification A macro for Excel was written to filter the data and identify instances of purposeful, directional travel (either upstream or downstream) during each 24-hour period. Due to the possibility that a fish could travel between the two antennas by either a short route (i.e. through the narrow portion of the tank where the antennas are 0.5 m apart) or a long route 54 (i.e. beginning at one antenna and swimming around the perimeter of the inner baffle to the second) it was necessary to define directional movement as movement past the two antennas by the shortest distance (i.e. 0.5 m). A threshold of 4 seconds between the antennas detecting the same tag was used to define purposeful movement since to swim around the perimeter of the inner baffle (a distance of 2.1 m) in 4 sec, a 15 cm fish would have to swim approximately 3.5 body lengths per second. With a similar experimental design, Fangstam et al. (1993) reported maximum migration rates of 2.3 body lengths per second for 15 to 19 cm Baltic salmon (Salmo salar) (a linear distance of 1.4 to 1.7 m). Therefore, it would not be possible for a fish to be detected by antenna #1, swim around the perimeter of the inner baffle, and then be detected by antenna #2 in less than 4 seconds. As a result, purposeful movements were defined as those where two different antennas recorded the same tag in less than 4 seconds. Therefore, readings separated by greater than 4 s were removed from the dataset as they were assumed to represent either movement between the two antennas around the wider portion of the tank or slow non-directed movement past the antennas in the narrow portion. In addition, readings separated by less than 1 second were considered cross-talk between the antennas and were also removed. Net movements were calculated by subtracting upstream movements from downstream movements. Since the number of fish within each tank was reduced through sampling and mortalities during the course of the experiment, the daily movements were reported as a proportion of the number offish in the tank. Due to the disturbance within the 55 tank on the days of the physiological sampling and tank cleaning (which occurred on the same days), those days were not included in the dataset. Change-Point Analysis Movement data were analyzed using a change-point analysis. This technique is commonly used to assess whether a change has occurred in time ordered data and is based on comparing each observed value to the mean and identifying the points where there is a change from above average to below average (and vice versa). The technique is extremely flexible and robust to issues of non-normality and outliers within the dataset (Taylor 2000a). To apply this technique to a dataset the only assumption that needs to be met is that of independence. Bootstrapping (10,000 iterations) was used to provide an estimated confidence level that a detected change did occur and only changes with 99% or greater confidence were reported. Confidence intervals that define the time period of the change were also generated and set at 95%. All change-point calculations were completed using the commercial software Change-Point Analyzer (ver. 2.3, Taylor 2000b). Physiological Sampling During the tagging event in both years, a random selection of fish (2005 n = 9; 2006 n = 12) were sampled to provide baseline physiological measurements prior to the smolting period (mid to late March). Subsequent physiological sampling was then conducted at twoweek intervals beginning in mid-April. During each sampling event, 6 to 8 fish from each tank were randomly selected and euthanized in a solution of MS-222 (200 mg/L) buffered 56 with sodium bicarbonate (pH = 7.0) and the length and weights were collected. The caudal fin was severed and blood was collected in capillary tubes. Plasma was then separated from the whole blood by centrifugation (11,500 g for 3min). The time between capture and collection of blood was less than 5 minutes for each fish in order to limit the influences of stress on physiological samples. A gill biopsy of 2-4 primary gill filaments was also taken for analysis of gill Na+,K+-ATPase activity. Filaments were placed in individual tubes along with 100 ul of ice-cold SEI buffer (150 mM sucrose, 10 mM Na2EDTA, 50 mM imidazole, pH=7.3). All samples were frozen to -20°C within 1 hour of collection and transferred to a -80°C freezer within 12 hours of collection. Sampling was completed in both 2005 and 2006 for each tank with the exception of tank 2, for which only 2005 samples were collected due to mortalities in 2006. Specific sampling dates (2005/2006) were: March 15, April 22/21, May 7/6, May 20/22, June 3/3, June 17/17, and June 29/July5. Physiological Data Analysis I measured gill Na+,K+-ATPase activity using the microassay method of McCormick (1993). I quantified plasma Cortisol concentrations using a competitive solid-phase microtitre enzyme immunoassay (EIA) following the protocol outlined by Carey and McCormick (1998). In order to increase the sample size for each sampling event, data from 2005 and 2006 were pooled. A variance ratio test was used to compare the standard deviations from each year to test the assumption of equal variance. In cases where variances from the two years were found to be unequal (Na+,K+-ATPase activity n = 6; plasma Cortisol 57 concentration n = 13) a t-test assuming unequal variance was used. The remaining cases where variances did not differ significantly (Na+,K+-ATPase activity n = 13; plasma Cortisol concentration n=6) were compared with a two sample t-test. For Na+,K+-ATPase activity significant differences between the two years of sampling were detected for 3 of the 19 comparisons: April 22/21 for tank 1 and 3, and June 3 for tank 3. For plasma Cortisol concentrations significant differences between data collected in 2005 and 2006 were detected in 5 of the 19 comparisons: May 7/6 for tank 1; June 3 for tank 4; June 29/July 5 for tanks 1, 3 and 4. Without an additional year of sampling it is not possible to determine which year's data is more accurate in each of those instances and in order to increase sample size and statistical power the decision was made to pool the data despite the differences. For the remaining sampling event, no significant differences (a = 0.05) were detected. Problems Encountered Over the course of the two years of the experiment, several minor problems were encountered including brief fluctuations in temperature regime, occasional mortalities, and short periods (i.e. less than 1 week) where the PIT tag readers were not working. None of these problems were considered significant enough to change the experimental design or significantly impact the results. However, there were two notable problems which limited that data collected and hence the ability to make inferences. First, during the initial run of the experiment (2005) technical problems associated with one of the PIT tag readers prevented the collection of migration data from tanks 3 and 4 (constant temperature with 58 high and low flow, respectively). Unfortunately, by the time the repairs to the unit were completed by the manufacturer, the first run of the experiment was completed. The second notable problem occurred during the second run of the experiment in mid-April when all fish in tank 2 died. The fish died when they were removed from the tank and placed in a holding bucket, while the tank was being cleaned. The cause of the mortalities is not known but normal cleaning procedure did not involve removing the fish from the tank, however, in this particular case the fish were removed while the tank was drained. As a result of this event, only one year of data was available for tank 2. Despite these problems, I felt that there was adequate data collected to proceed with the analyses and although it would have been useful to include an additional year's data, I did not consider it critical given the results obtained. In addition, tank 1, which was the only tank for which 2 years of data were collected for all parameters, showed a consistent movement pattern in both years suggesting that the data collected for the remaining 3 tanks in only year was likely representative of the behavioural response. Data Analysis The results of the physiological sampling were compared using a 3 factor ANOVA to detect differences associated with sampling event (7 levels), temperature (2 levels) and flow (2 levels). All possible interactions of the 3 factors were also included in the model. Prior to the ANOVA, a Shapiro-Wilk test was used to test for normality for each of the 25 pooled datasets. Both the ATPase and Cortisol data contained several instances of non-normality 59 and therefore a ladder of powers analysis was used to identify a suitable data transformation to minimize non-normality. For the ATPase data a square root transformation was used while for the Cortisol data a log transformation was used. A Bartlett's Test was used to test for equality of variance after each comparison. Post-ANOVA comparisons were completed using a Tukey multiple comparison test. An alpha value of 0.05 was used for all tests. Modeling: Physiology and Behaviour Interaction In order to combine the environmental, physiological, and migration data, an ITMC approach using a count model was used. A detailed discussion of this technique along with the procedure for implementation is given in Chapter 1. The dependent variable used for the analysis was daily movements within each experimental tank expressed as a proportion of the total movements within the tank over the course of the study. Data from year 1 and 2 were combined for the modeling analysis. The environmental, physiological and descriptive parameters included in the ITMC analysis are provided in Table 6. 60 Table 6. Parameters used in the ITMC analysis of Chinook smolt downstream migration (daily proportion of total movements) during a controlled experiment. Parameter Description Squared daily mean temperature. Squared cumulative daily mean temperature to describe temperature experience. Flow (m/s) Flow velocity within tank. Photoperiod (hours) Period from sunrise to sunset. Length (mm) Mean length of fish sampled during each sampling event. Weight (g) Mean weight of fish sampled during each sampling event. Mean condition factor of fish sampled during Condition factor each sampling event. Mean Na+,K+-ATPase activity of samples ATPase collected during each sampling event. Mean plasma Cortisol of samples collected Cortisol during each sampling event. * Where a squared term is included, the linear term is included as well. Temperature (°C)* Accumulated Thermal Unit2 (ATU)* As a result of sampling at two-week intervals, daily data were not available for the physiology parameters (gill Na+,K+-ATPase activity and plasma Cortisol) or the descriptive parameters (length, weight, condition factor). In order to approximate daily values for these parameters for use in the model, the mean of all samples collected at each of the two-week sampling intervals was calculated and that value was used for the week prior to and immediately following each sampling event. A total of 30 models were developed using the 9 parameters outlined in Table 6 (see Table 7 for list of models). Due to overdispersion in the dataset, the zero-inflated negative binomial (ZINB) model was preferred over the zeroinflated poisson (ZIP) model (G = 3563.475, p < 0.001). Similarly the ZINB was preferred 61 over the negative binomial regression model based on the results of a Vuong test (V = 6.190, p < 0.001). The inflation term ATU2 was used in each of the models. AICC values were calculated for each and the models were ranked based on that value. Results Temperature Temperature increased steadily in tanks 1 and 2 over the course of the experiment (Figure 9). Initial temperatures of approximately 4°C (2005) and 6°C (2006) increased to between 12 and 14°C by the end of the experiment. Tanks 3 and 4 were kept below 6°C, throughout the experiment for both years with the exception of the period between the 2nd and 7th of May, 2005, when the temperatures exceeded 6°C due to an unanticipated spike in the river temperature. A two-sample t-test was used to confirm there was no significant difference in temperature between the two years of the experiment for Tanks 1 and 2 (p = 0.43) and Tanks 3 and 4 (p = 0.89). 62 erature (C) 10 15 Tanks 1 and 2 f o 0.60W ;2 0.40 Q 0.20 _>> 0.00-0.20 -0.40 § -0.60 to (55> a 2 cD 3, ( *m >r .*..*.* J$ 1 *n <3\ CN i~< J-H QH 9« < —< 99% confidence; 10 000 bootstraps). A total of seven change points were identified for tank 1. The first identified change was an increase in migrations on Apr. 15 followed by a second increase 8 days later on Apr. 67 23 r . Migrations then remained fairly constant until May 21 s when there was another identified increase followed by a decrease 1 week later (May 28th). This cycle of increased movement for a week followed by a decrease for a week continued for the remainder of the experiment. Tank 2 showed only two change-points with an increase in downstream migrations on Apr. 24th and a decrease on May 29th. Three change-points were identified for tank 3 with an initial increase on May 4th followed by second increase on May 17th. The third point (June 4 ) was identified at the point were movements began to decrease. No significant change-points that met the criteria of the analysis were identified for tank 4. A general increase in downstream migrations was observed, however, it was a steady increase as opposed to an abrupt change. Physiology Gill Na+X-ATPase Activity The Shapiro-Wilk test for normality identified 8 (29%) cases of non-normally distributed data before transformation. The results of the ladder of powers analysis identified the square root transformation as being most appropriate for the data reducing the cases of non-normality to 3 (11%). Tanks 1 and 2, which had an increasing temperature regime showed a peak in gill Na+,K+-ATPase activity in mid-May, whereas tanks 3 and 4, which had a constant 68 temperature regime, showed a steady increase in activity but no peak over the study period (Figure 12). Increasing Temp, and Low Flow 10- n_ Q _ 6- mg pro!tei € s < • o I 4as t < 2" M arch 1 * Increasing Temp, and High Flow € 10c -2 "o a. 8 - hI < April 1 . 6- 1 o I I • 1 48! ro P: 2 - l June 1 May 1 } E E Q l July 1 < 1 April 1 l May 1 protei 00 •6 1 0 c "3 "o Q. 8 D) E 6- o • *• TPase 1 < l July 1 Constant Temp, and Low Flow 10- U) < l June 1 Date Constant Temp, and High Flow D I • l March 1 Date € c * i i , i I . T * I April 1 0 < 0 to to E May 1 5* 1 4- I• i June 1 6- o m i M arch 1 E i July 1 < 21 March 1 " 5 1 April 1 1 * * 1 May 1 1 June 1 1 July 1 Date Date Figure 12. Average gill Na+,K+-ATPase activity (umol ADP/mg protein/h) values (± SE) from samples collected from Chinook smolts (Oncorhynchus tshawytschwd) in 2005 and 2006 from four experimental tanks. Square and diamond data points represent yearly averages (2005 and 2006, respectively) for sampling events where a significant difference was detected between the two years. Results of the 3-factor ANOVA (Appendix 2d) showed that temperature and sampling event had a significant effect on gill Na+, K+-ATPase activity (p < 0.001) while 69 flow did not (p = 0.84). The interactions "temperature + event", "flow + event", and "temperature + flow" were found to be significant (p < 0.001, p = 0.012, and p = 0.011, respectively). The significance of these interactions was primarily attributed to the influence of temperature and event on the enzyme activity given that each interaction with flow was less strongly significant than when flow was not included. The final interaction "temperature + flow + event" was found to be strongly non-significant (p - 0.68). Fish in each of the tanks showed an increase in gill Na+,K+-ATPase activity over the course of the experiment and significant differences between the initial enzyme activity (March 15th) and that of subsequent sampling events in all 4 tanks were noted. However, significant increases from the initial value were detected approximately 2 weeks earlier in the warm tanks (1 and 2) than in the cold tanks (3 and 4). The effect of increased temperature on Na+,K+-ATPase activity was made apparent by the comparison of activities at specific sampling events which identified significantly higher activity at sampling events 3 (May 7th) and 4 (May 20th) when each of the warm tanks (1 and 2) were compared to the cold tanks (3 and 4). However, events 5 - 7 (June 3rd, 17th, and July 6th) were not significantly different since Na+,K+-ATPase activity in warm tanks had begun to decline during that period whereas in the cold tanks it was still increasing. No significant differences in Na+,K+-ATPase activity were identified at any of the sampling events when the low and high velocity tanks were compared either at warm temperatures (1 vs. 2) or low temperature (3 vs. 4), confirming flow velocity did not influence enzyme activity. Tank 1 (increasing 70 temperature, low velocity flow) was the only tank to show a significant change in Na ,K ATPase activity between subsequent sampling events. This occurred when activity significantly increased between event 3 (May 7th) and 4 (May 20th). The remaining tanks showed non-significant changes between sampling events. Plasma Cortisol Plasma Cortisol levels for tanks 1 and 2 (temperature increase) showed a decline from the initial values (i.e. March 15th) at sampling events 2 through 6 (April 22 to June 17 ) and did not increase past the initial level until sampling event 7 (July 5th) for either tank (Figure 13). Alternatively, Cortisol levels in tanks 3 and 4 (constant low temperature) remained at a level similar to that of the initial sampling event until declining at event 4 and 3, respectively, and then remained below the initial value. 71 Increasing Temp and Low Flow Increasing Temp and High Flow 200- E °> 150 H 150- K 100 H 100- I * " i f i 50- i i 0- March 1 April 1 May 1 June 1 July 1 i i March 1 April 1 i May 1 Date i July 1 Date Constant Temp, and High Flow Constant Temp, and Low Flow 200- ml) 200- 1 °> ISO- |> ISO- 's 'S 't 100- * 100o ra * o Plasma C i June 1 o- { i i March 1 April 1 i i May 1 . i * i June 1 5 I 1 50- . 0i— July 1 n 1 March Date I r 1 April i 1 May - 1 - • • 11 June i July 1 Date Figure 13. Average plasma Cortisol (ng/ml) values (± SE) from samples collected from Chinook smolts (Oncorhynchus tshawytschwa) in 2005 and 2006 from four experimental tanks. Square and diamond data points represent yearly averages (2005 and 2006, repectively) for sampling events where a significant difference was detected between the two years. The results of the 3-factor ANOVA (Appendix 2e) showed that sampling event, temperature and flow all had significant effects on plasma Cortisol (p < 0.001). In addition, all of the interactions with the exception of "temperature + flow" (p = 0.38) were also found to be significant (p < 0.02). In general, the tanks that experienced high velocity flow (tanks 2 72 and 3) showed less variability in Cortisol levels which remained similar to the initial values for longer. Fish from tank 2 were only found to have Cortisol levels significantly higher than the initial value at sampling event 7 (p = 0.044), while in tank 3 differences were identified at events 4 and 5 (p = 0.001 for both). Alternatively, the tanks with no flow manipulation (tanks 1 and 4) tended to decline from the initial value earlier and remain at low levels (with the exception of sampling event 7 for tank 1). Both tanks 1 and 2, which experienced increasing temperature, showed a significant increase in plasma Cortisol between sampling event 6 and 7 (p < 0.001). This increase is not considered due to sampling induced stress since a similar increase was not identified in tanks 3 and 4 suggesting it may be a temperature related response. Modeling Based on AICC scores, the model that best described the movement data collected was a combination of ATP and photoperiod. However, the model ranked 2nd (ATP + ATU) had an AICC scores within 2 suggesting a similar ability to describe the observed movement patterns (Table 7). As a result, ATU and photoperiod are almost interchangeable when combined with ATP. The model ranked 3rd also had an AICC score within 2, however, since this model included one additional parameter it was considered a less parsimonious model. 73 Table 7. Summary of model selection statistics for the candidate models to describe the downstream migration pattern of juvenile Chinook salmon {Oncorhynchus tshawytschd) in a controlled laboratory experiment. Zero-Inflated Neg. Binomial Model ATP + P ATP + A2 ATP + F + P L + A2 + ATP CF + A2 + ATP ATP + A2 + P L + A2 + ATP + P A2 + F + P + ATP W + A2 + ATP + P A2 + F + P + L + W + CF + ATP + C L + A2 + ATP + P + F CF+ A2 + ATP + P + F W + A2 + ATP + P + F C + A2 + P L + A2 + P C + A2 + F + P A2 W + A2 + P A2 + F + P L + A2 T2 + F + P A2 + F W + A2 ATP C + T2 + F +P ATP + C ATP + F C+F+P L + W + CF T2 + F Rank 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 AICC 2762.422 2763.005 2764.133 2764.500 2764.772 2765.012 2766.469 2767.027 2767.065 2767.233 2768.566 2768.774 2769.104 2772.038 2773.056 2773.079 2773.158 2773.838 2773.950 2774.391 2774.630 2774.671 2774.865 2776.102 2776.193 2776.292 2777.526 2784.397 2802.321 2808.843 AAICC 0.0 0.6 1.7 2.1 2.4 2.6 4.0 4.6 4.6 4.8 6.1 6.4 6.7 9.6 10.6 10.7 10.7 11.4 11.5 12.0 12.2 12.2 12.4 13.7 13.8 13.9 15.1 22.0 39.9 46.4 A2 = ATU quadratic; ATP = ATPase; C = Cortisol; CF = Condition factor; F = Flow L = Length; P = Photoperiod; T2 = temperature quadratic; W = Weight 74 Each of the top 13 models in the analysis included the parameter ATP highlighting its correlation to migration. Similarly, ATU was included in 8 of the top 10 models while temperature was not included until the model ranked 21st. The parameter flow was included in the model ranked 3rd, but only appears 3 times in the top 10 models. Similarly, the parameters condition factor and length were included in only 2 and 3 of the top 10 models, respectively. Parameters that did not show a strong correlation with migration included Cortisol and weight, each of which only appeared once in the top 10 models. Neither ATU nor ATP were found to be strongly correlated to the observed movement patterns when considered individually (rank 17 and 24, respectively) suggesting a combination of parameters is involved. 75 Discussion The goal of this study was to investigate the role of temperature and flow velocity on the migration timing of Chinook salmon smolts and determine if this timing was associated with an increase in gill Na+,K+-ATPase activity using a controlled laboratory experiment. In particular, a warmer temperature regime was found to accelerate both behavioural and physiological characteristics of smolting. The onset of movement as well as the decline in movement was found to occur earlier in the experimental tanks that experienced increasing temperature. Likewise, the increase in gill Na+,K+-ATPase activity at warmer temperatures was advanced. The influence of flow on smolting was less obvious. Movement data suggests a possible role of flow in regulating the migration process since the tanks that experienced high flow velocities had a more well-defined migration period (i.e. clear increase, peak and decline in movement) when compared to those without the influence of flow. However, the results of the physiological analysis show no influence of flow on gill Na+,K+-ATPase activity or plasma Cortisol concentration suggesting flow is likely not a critical component in initiating the physiological changes associated with smolting. Therefore, while not involved in the onset of migration, flow may play a role in controlling the migration pattern once underway. This was supported by the results of the modeling analysis which identified a combination of gill Na+,K+-ATPase activity with either photoperiod or ATU as being the most important variables in describing the observed movement patterns within the four experimental tanks. 76 The use of a controlled laboratory experiment to isolate the roles of individual environmental parameters on smolting is relatively novel with the bulk of existing work focusing on observations of the migration patterns of wild populations. The current experiment was based on that of Zydlewski et al. (2005) and Fangstam et al. (1993), and was expanded to investigate the role of flow velocity in addition to temperature and focuses on Chinook as opposed to Atlantic salmon. In regards to the role of temperature on migration, my study supports the findings of both Fangstam et al. (1993) and Zydlewski et al. (2005). In particular, fish experiencing an increasing temperature regime showed an earlier increase in movement than fish under constant temperature. Physiological sampling confirmed that the increase in movement corresponded to an earlier increase in gill Na+,K+-ATPase activity and therefore was associated with smolting as opposed to behavioural interactions within the tanks. The physiological response observed was consistent with previous studies that reported an accelerated temperature increase resulted in an earlier increase in gill Na+,K+ATPase activity (McCormick et al. 1997). I did not find a concomitant increase in plasma Cortisol levels with increases in gill Na+,K+-ATPase activity as has previously been shown in smolting salmon (Shrimpton and McCormick 1998). There was little effect of the experimental regime on plasma Cortisol, except for the last sampling point; likely due to the warmer temperatures. Seasonal increases in hormones associated with smolting have been repeatedly shown (for reviews see McCormick et al. 1987; Hoar 1988), but hormonal surges may differ between years within the same population of fish (Shrimpton et al. 2000), which could have contributed to the lack of observed increase. The results could also be interpreted 77 to mean that physiological changes associated with smolting are stimulated without substantial increases in all hormones, perhaps due to synergism between hormones such as Cortisol or growth hormone (McCormick 1995; Shrimpton et al. 1995). As sufficient plasma was not collected to analyze growth hormone, it is unclear what affect the experimental regime had on this hormone. My study also showed that loss of physiological and behavioural characteristics associated with smolting was related to temperature regime suggesting that smolting is reversible. The loss of physiological characteristics was only observed in the experimental tanks that experienced increasing temperature while the tanks that experienced a constant temperature showed a relatively steady increase in gill Na+,K+-ATPase activity with no obvious peak or subsequent decline. Despite the similar physiological response of the constant temperature tanks, the behavioural response in those two tanks varied. For example, tank 3 (constant temperature, high velocity flow) showed a similar pattern to that of tank 2 (increasing temperature and high velocity flow) albeit with an increase and decline in movement that was delayed by 1-2 weeks as a result of the difference in temperature. Tank 4 (constant temperature, low velocity flow), on the other hand, never did show a movement pattern indicative of downstream migration. Based on these results it is apparent that temperature alone is not sufficient to explain observed movement patterns. Instead flow may be contributing to the ultimate movement patterns observed. This is further supported by movement data from tank 1. Fish in this tank, which experienced an increase in temperature 78 but low flow, displayed an abnormal movement pattern (Figure 10) characterized by a series of initial increases in movement followed by a cyclical increase-decrease pattern at approximately 10-day intervals. My results also showed that temperature experience (ATU) correlated more strongly with the initiation and termination of downstream movement than daily mean temperature, suggesting that a threshold temperature is not as important factor for movement and supports the work of Zydlewski et al. (2005). Further support for the importance of temperature experience is apparent from the modeling analysis, which included ATU in 8 of the top 10 models while temperature was not included until the model ranked #21. Further, the importance of temperature experience has also been linked to smolt physiology and in particular the increase and subsequent decrease of gill Na+,K+-ATPase activity (McCormick et al. 1997). The springtime increase in water temperature (tanks 1 and 2) resulted in peak gill Na+,K+-ATPase activity in late May followed by a decline. The constant temperature groups (tanks 3 and 4) showed a steady increase in enzyme activity until the end of the study in early July which corresponds to the end of the smolt period for Chinook salmon from the Upper Fraser. At the end of the experiment, ATU of 500 were reached for the cold groups (tanks 3 and 4) and gill Na+,K+-ATPase activities were approximately 6 umol ADP/mg protein/h. However, in the warmer tanks both those levels had been achieved by mid-May. If the experiment had been extended, it is not known if smolt characteristics (based on enzyme profiles) would have continued to develop. It is likely, however, that behavioural 79 and physiological characteristics associated with smolting in the colder tanks would not likely have achieved the same levels as seen in the warmer tanks (see Zaugg et al. 1972; Zaugg and McLain 1976; McCormick et al. 2000; McCormick et al. 2002). It is apparent that under low temperature conditions, a delay in physiological development could result in fish missing the physiological smolt window. Previous studies have shown smolting will not occur in the absence of a photoperiod signal and temperature alone cannot initiate the physiological changes associated with smolting, but instead controls the rate of change (Muir et al. 1994; McCormick et al. 2002). The results of this study show that flow has no effect on the rate of physiological change such that increased flow velocity will not advance smolting in the absence of a temperature increase. However, in regards to migration, the results suggest that while the parameter does not appear to be involved in the initiation of migration as has been suggested (McCormick et al. 1998), it may be important to controlling the pattern of migration once underway. Without the input of strong directional flow, the fish appear to either respond with cyclical pulses of movement (for example tank 1 temperature increase with no flow increase; Figure 11) or no defined movement pattern (for example tank 4 no temperature or flow increase; Figure 11). However, when strong directional flow was included the movement pattern was continuous and well-defined (tanks 2 and 3, Figures 11). One possible explanation for this is that the presence of strong directional flow helps orient the fish downstream and increases the overall rate of travel resulting in a more coordinated movement pattern. In addition, 80 studies have shown that the migration process itself has been linked to the continued development of smolt characteristics. In particular, Zaugg et al. (1985) found that fish that migrated had 2.5 times greater gill Na+, K+-ATPase activity than those that were maintained in a hatchery, suggesting migration promotes the development of seawater tolerance. This relationship could be viewed as a feedback loop where the behavioural response (migration) leads to increased physiological response which in turn leads to increased behavioural response (and so on). For populations with long migratory routes which migrate earlier in the season, the physiological changes taking place during migration are considered important to subsequent survival in the marine environment. Therefore, factors such as flow that influence the migration pattern must be considered when management decisions are being made. It has been suggested that flow augmentation during the migration period can be used as a management strategy to help ensure smolt survival (Connor et al. 2003). The main rationale for this strategy is the belief that augmented flows increase the rate of seaward movement of smolts thus ensuring more individuals reach the marine environment. The results of this work suggest that flow augmentation would also result in a more coordinated migration event. However, any manipulation of flow would need to be carefully timed to ensure it was synchronized with the natural hydrograph of the system as well as the changes in other environmental variables such as photoperiod and temperature. For example, were the augmentation to occur too early (i.e. before the increase in photoperiod initiates the 81 smolting process), the fish would not yet be ready to migrate and therefore would not benefit from the augmentation. Similarly, flow augmentation that occurs too late might be of little or no benefit to fish that have already moved downstream. Another consideration is that flow augmentation could result in unanticipated manipulation of other parameters such as temperature and turbidity (Connor et al. 2003). For example, depending on whether the water being added to the system was drawn from the top of a reservoir where it is warmer or the bottom where it is colder, there could be a substantial change in temperature associated with the increased flow. Unusually warm conditions could result in accelerated development and loss of smolt characteristics whereas unusually cold conditions could delay the development and loss of smolt characteristics. Therefore, while flow augmentation has the potential of benefiting migrating smolts, the timing and duration of the increase as well as its effect on other parameters must be considered. The results of the modeling analysis, in general, do not emphasize the role of flow to the observed movement patterns. The parameter was only included in 3 of the top 10 models with greater than 50% of the models that included flow being ranked 15 or higher. Flow was included in the model ranked 3rd; however, an analysis of the role of each parameter in that model showed that flow had a non-significant (p = 0.85) influence whereas the other parameters included in the model (ATP and photoperiod) had significant effects (p < 0.001). In addition, the version of that same model without the flow parameter was ranked first, meaning that model did a better job of describing the observed movement patterns without 82 the inclusion of the flow parameter. It is important to note that this experiment focused on flow in terms of velocity of water and not discharge volume, which would also be expected to increase in the spring. Due to limitations with the water supply at the Quesnel River Research Center it was not possible to substantially increase the volume of water to 2 of the 4 tanks without potentially impacting the remaining tanks. As a result, the decision was made to keep a constant volume of water flowing into each tank but adjust the angle of the spray bars to achieve increased velocities in tanks 2 and 3. A similar experiment that included an increase in discharge volume and rate may identify discharge as being more important to the overall migration process. This was the case in the modeling analysis of historical data from the Nechako River, which identified a combination of ATU and flow (in that analysis flow was equivalent to discharge) as being most correlated to the observed migration patterns. Similarly, limitations with the experimental setup did not make it possible for flow to be increased over the course of the experiment. Instead, it was maintained at either a constant high or low velocity for the duration of the experiment. As a result, there was no increase or decrease in velocity to which changes in movement could be correlated. Despite this, the change-point analysis showed that flow likely is a factor in controlling migration patterns and it is likely that a similar experimental setup which allowed for natural increases and decreases in flow would identify a stronger correlation between flow and migration. 83 General Summary 84 My thesis analyzed the role of selected environmental parameters on the migration timing of Chinook salmon smolts using both observational and experimental methods. In the observational study (Chapter 1), I used a statistical modeling technique to identify correlations between 13 years of Chinook smolt migration data from the Nechako River with changes in environmental parameters. In general, the migration pattern of smolts from the Nechako system correlated to increasing ATU. This suggests that while decreased spring temperatures would artificially extend the smolting period due to slowing down the rate of physiological change, warmer than normal conditions would have the opposite effect. This was confirmed by examination of the historical data which showed that in warmer years fish tended to migrate earlier from the system than in cooler years. In the experimental study (Chapter 2), collection of blood and tissue samples from fish allowed for the inclusion of physiological parameters (gill Na+,K+-ATPase activity, plasma Cortisol concentration) along with physical measurements (length, weight, condition factor) and environmental parameters (temperature and flow). Temperature was found to accelerate the physiological changes associated with smolting such that gill Na+,K+-ATPase activity data showed a period of increase followed by a peak and decline. Fish in the tanks that did not experience the temperature increase showed a steady increase in gill Na+,K+-ATPase over the length of the experiment with no defined peak. Length, weight and condition factor of fish in the increased temperature treatments were found to be significantly higher than that of the fish in the constant temperature treatments. The data on movement pattern from the experimental tanks also showed that changes tended to occur earlier in those tanks that experienced a 85 temperature increase. These findings support the predictions from Chapter 1 that earlier increases in temperature result in earlier readiness to migrate. The influence of flow on migration was less well defined. The model analysis completed in Chapter 1 identified flow as being correlated to observed migration patterns. However, the negative value assigned to the flow parameter suggests a negative influence on migration such that an early peak in flow would result in an earlier termination of migration. The possibility of flow being a cue for the termination of migration has not been previously suggested in the literature and further study would be required to investigate this possibility. In addition, several studies have previously suggested that flow augmentation in the spring can be beneficial to migrating smolts, which would seem to be contradicted by the model results. A review of the historical data (Figure 3) shows that fish from the Nechako River typically begin to migrate prior to the increase in flows in the spring and that the flow increase actually corresponds to the period when the migration is coming to an end. Therefore, the results of the model are accurate in the context of the data and the particular hydrograph of the Nechako River and it is important that the results are not misinterpreted to suggest that increasing flow has a negative influence on migration. While it is likely that photoperiod and temperature are more involved in the onset of smolting, flow likely does play an important role once migration is underway. This was supported by the experimental analysis which showed that flow does not have a noticeable effect on the physiological changes associated with smolting or the onset of migration. However, the experimental 86 tanks that did not experience an increase in flow showed either irregular pulses of migration (tank 1 - temperature increase) or no increase in migration (tank 4 - no temperature) suggesting that flow does play an important role in controlling or modulating the migration pattern once underway. Although the experimental design was limited to examining flow velocity, I believe the results do provide a reasonable indication of the role of discharge on the migration process. However, it would be interesting to examine if accelerated increases in discharge volume and velocity would have more of an influence on the physiological aspects of smolting as well as the onset of migration. A comparison of the two models (Chapter 1 - historical data; Chapter 2 experimental data) shows several similarities and some contradictions. Both models identified the importance of temperature, specifically ATU, on migration. However, while the selected models based on the historical data included flow, the experimental model did not. This contradiction can be attributed to the differences in the flow data in each model. First, as discussed in the previous paragraph, the use of flow discharge (historical model) and flow velocity (experimental model) could potentially account for the difference. However, since increases in velocity and increases in volume in the spring are correlated to a certain degree, it seems reasonable that changes attributed to one would be representative of the other. The second, and perhaps more important difference, was that in the historical dataset flow showed both an increase and decrease during the outmigration period, while the experimental analysis kept flow at a constant level (either high or low) for the duration of the 87 experiment. This difference could partially explain the difference in relative importance of flow in each analysis. The change-point analysis provided an additional tool to assess the role of flow which was shown to be more in agreement with the results of the historical data analysis as well as existing literature. I believe that if the experiment were repeated with flow manipulated to have an increase and decrease, there would be a stronger correlation to movement similar to what was observed with the historical data. Another slight contradiction between the two models was the importance of photoperiod. The historical analysis did not include photoperiod in either of the selected binary or count models, while it was included in the selected model in the experimental analysis. A review of the overall rankings of models in the historical analysis showed that for both the binary and count analyses photoperiod was included in the model ranked second. For the binary analysis the model with photoperiod was actually ranked higher based on AICC score, but the model ranked second was chosen since it included one less parameter. Therefore, it can be argued that both modeling analyses identify the overall importance of photoperiod to the smolting process and are in agreement with the research that has been completed in that area. The fact that models without photoperiod performed nearly as well or better, does suggest photoperiod may be less involved once migration is underway. The final difference between the two models is that for the experimental analysis, gill Na+,K+ATPase activity was included and the results showed a strong correlation between movement and that parameter. This result is in agreement with existing work that has been completed 88 on the smolting process and the role of gill Na ,K -ATPase as an indicator of readiness to enter seawater. Although this parameter was not included in the historical model, the results of the samples collected from migrating smolts in the Nechako River in 2004 (Figure 1) show a similar trend to what was observed in the experimental analysis for tank 2, which experienced increasing temperature and high flow (Figure 12). Based on the results of this study, it seems apparent that flow manipulations that change the timing, duration or magnitude of increases of temperature and flow in the spring could potentially have adverse effects on migration. Both analyses identified temperature experience (ATU) as being more strongly linked to migration than a daily or threshold temperature. In addition, both showed that earlier, warmer temperatures will result in earlier migration. Alternatively, while flow may not be a cue for the onset of migration, it does potentially play an important role once migration is underway and may even serve as a termination cue in some systems. Therefore, it is important to consider both parameters when making management decisions. For example, flow augmentation for the purpose of accelerating migration rate could result in either an increase or decrease in temperature depending on the source of the water added to the system. This, in turn, could influence the rate of physiological change such that the fish still do not arrive at the marine environment when changes are optimal. Another situation where caution must be taken is when one variable in a system is manipulated independent of the others. Within the Nechako system, construction of a cold-water release facility at the Kenny Dam has been proposed as a means 89 of controlling high summer water temperatures. This facility would use a small volume of cold water from the bottom of the reservoir to reduce summer water temperatures in the river during the spawning migration of sockeye salmon. In the past, attempts at temperature manipulation have relied on water being released from the surface of the reservoir which, due to being only slightly cooler than the river water, requires a substantial volume of water to be released over a long period of time. This results in a large increase in flow for a minimal decrease in temperature whereas the proposed facility would allow for a substantial change in temperature while minimizing the change in flow. While the ability of managers to decrease temperature without affecting flows might seem like a good strategy to address high summer temperatures, the results of this study and others have shown that most ecological processes are controlled by more than one variable. As a result, it is unclear what effect manipulation of single variable in a suite of control variables will have. Due to the current management efforts for the Nechako River focusing on sockeye migration in the fall, manipulation of the flow regime does not generally occur during the spring and as a result does not affect the Chinook smolting period. However, in a managed system unexpected events can occur such as emergency spilling when reservoirs are too full. Such a situation occurred in the Nechako River in the spring of 2007 when emergency spilling was necessary beginning in early March and continuing through the end of August. While it is unclear exactly what effect this may have had on smolt migration, the results of my study suggest that due to the decrease in mean daily temperature of approximately 4°C as 90 compared to 20-year average (Water Survey of Canada) associated with the increased discharge volume, the rate of physiological smolt development was likely slowed resulting in a delay in migration. Further, it is unlikely that the increase in discharge itself would result in an earlier smolting response since the required photoperiod cues would not have been received. However, it is possible that fish were forced out of the system early as a result of flows which were seven times higher than normal and which began two months before the normal onset of migration. Despite the potential impact of the 2007 emergency spilling, flooding is a natural process that occurs both irregularly and infrequently. As a result, it cannot be expected to have long-term effects on the overall health of the population. However, the same can not necessarily be said of flow regimes that are manipulated continuously without a solid understanding of the potential effects of those manipulations. My study shows the importance of considering both temperature and flow when making management decisions along with the timing of when manipulations should occur. However, due to genetic diversity, variable life-history strategies and plasticity of species and populations, as well as differences in flow and temperature regimes, and migration distances between systems, it should not be expected that each population will share the same suite of environmental cues. In the end, population-specific studies will still be required to assess the potential effects of proposed management strategies and both the ITMC modeling technique and experimental design used in this study are tools that can be used for that purpose. 91 References Anderson, D.R, Burnham, K.P., and Thompson, W.L. 2000. Null hypothesis testing: Problems, prevalence and an alternative. J. Wildl. Manage. 64: 912-923. Antonsson, T., and Gugjonsson, S. 2002. Variability in the timing and characteristics of Atlantic salmon smolt in Icelandic rivers. Trans. Am. Fish. Soc. 131: 643-655. Baker, P.F. and Morhardt, J.E. 2001. Survival of chinook salmon smolts in the SacramentoSan Joaquin Delta and Pacific Ocean. In Fish Bulletin 179: Contribution to the biology of Central Valley Salmonids. Vol. 2 Edited by Brown, R.L. Sacramento, Ca, California Department of Fish and Game. Pp. 163-182. Beckman, B.R., Larsen, D.A., Sharpe, C, Lee-Pawlak, B., Schreck, C.B., and Dickhoff, W.W. 2000. Physiological status of naturally reared juvenile spring chinook salmon in the Yakima River: seasonal dynamics and changes associated with smolting. Trans. Am. Fish. Soc. 129: 727-753. Boyce, M.S., Vernier, P.R., Nielsen, S.E., and Schmiegelow, F.K.A. 2002. Evaluating resource selection functions. Ecol. Model. 157: 281-300. Burnham, K.P., and Anderson, D.R. 2004. Multimodel Inference: Understanding AIC and BIC in model selection. Socio. Meth. Res. 33: 261-304. Carey, J.B., and McCormick, S.D., 1998. Atlantic salmon smolts are more responsive to handling and confinement stress than parr. Aquaculture 168: 237-253. Connor, W.P., Burge, H.L., Yearsley, J.R., and Bjornn, T.C. 2003. Influence of flow and temperature on survival of wild subyearling fall chinook salmon in Snake River. N. Am. J. Fish. Manage. 23: 362-375. Duston J., and Saunders, R.L. 1990. The entrainment role of photoperiod on hypoosmoregulatory and growth-related aspects of smolting in Atlantic salmon {Salmo salar). Can. J. Zool. 68: 707-715. Fangstam, H., Berglund, I., Sjoberg, M., and Ludqvist, H. 1993. Effects of size and sexual maturity on downstream migration during smolting in Baltic salmon {Salmo salar). J. Fish. Bio. 43: 517-529. 92 French, T.D., and Chambers, P.A. 1997. Reducing flows in the Nechako River (British Columbia, Canada): potential response of the macrophyte community. Can. J. Fish. Aquat. Sci. 54: 2247-2254. Guthery, F.S., Brennan, L.A., Petersen, M.J., and Lusk, J.J. 2005. Information theory in wildlife science: Critique and viewpoint. J. Wildl. Manage. 69: 457-465. Handeland, S.O., Wilkinson, E., Sveinsbo, B., McCormick, S.D., and Stefansson, S.O. 2004. Temperature influence on the development and loss of seawater tolerance in two fast-growing strains of Atlantic salmon. Aquaculture233: 513-529. Hvidsten, N.A., Jensen, A.J., Vivas, H., Bakke, O., and Heggeberget, T.G. 1995. Downstream migration of Atlantic salmon smolts in relation to water flow, water temperature, moon phase and social interaction. Nord. J. Freshw. Res. 70: 38-48. Hoar, W.S. 1988. The physiology of smolting salmonids. In Fish Physiology, vol 1 IB Edited by W.S. Hoar and D.J. Randall. Academic Press, London, pp. 275-343. Iwata, M. 1995. Downstream migratory behavior of salmonids and its relationship with Cortisol and thyroid hormones: a review. Aquaculture 135: 131-139. Jonsson, N. 1991. Influence of water flow, water temperature and light on fish migration in rivers. Nord. J. Freshw. Res. 66: 20-35. Long, J.S., and Freese, J. 2006. Regression Models for Categorically Dependent Variables Using Stata. 2nd Ed. College Station, TX: Stata Press. Lord, D., Washington, S.P., and Ivan, J.N. 2005. Poisson, Poisson-gamma and zero-inflated regression models of motor vehicle crashes: balancing statistical fit and theory. Accid. Anal. Prev. 37: 35-46. McCormick, S.D., Saunders, R.L., Henderson, E.B., and Harmon, P.R. 1987. Photoperiod control of parr-smolt transformation in Atlantic salmon (Salmo salar). Changes in salinity tolerance, gill Na+,K+ ATPase activity, and plasma thyroid hormones. Can. J. Fish. Aquat. Sci. 44: 1462-1468. McCormick, S.D. 1993. Methods for non-lethal gill biopsy and measurement of Na+, K+ATPase activity. Can. J. Fish. Aquat. Sci. 50: 656-658. McCormick, S.D. Hormonal control of gill Na+, K+-ATPase and chloride cell function. In Fish Physiology, Vol. XIV, Ionoregulation: cellular and molecular approaches. 93 Edited by C M . Wood and T.J. Shuttleworth. Academic Press, New York USA. Pp. 285-315. McCormick, S.D., Bjornsson, B.Th.,Sheridan, M., Eilertson, C , Carey, J.B., and O'Dea, M. 1995. Increased daylength stimulates plasma growth hormone and gill Na+, K+ATPase in Atlantic salmon (Salmo salar). J. Cell. Physiol. 165: 245-254. McCormick, S.D., Shrimpton, J.M., and Zydlewski, J.D. 1997. Temperature effects on osmoregulatory physiology of juvenile anadromous fish. In Global warming: Implications for freshwater and marine fish. Edited by C M . Wood and D.G. McDonald. Cambridge University Press, Cambridge UK. Pp. 279-301 McCormick, S.D., Hansen, L.P., Quinn, T.P., and Saunders, R.L. 1998. Movement, migration and smolting of Atlantic salmon (Salmo salar). Can. J. Fish. Aquat. Sci. 55 (suppl. 1): 71-92. McCormick, S.D., Cunjak, R.A., Dempson, B., O'Dea, M.F., and Carey, J.B. 1999. Temperature-related loss of smolt characteristics of Atlantic salmon (Salmo salar) in the wild. Can. J. Fish. Aquat. Sci. 56: 1649-1658. McCormick, S.D., Moriyama, S., and Bjornsson, B.Th. 2000. Low temperature limits the photoperiod control of smolting in Atlantic salmon through endocrine mechanisms. Am. J. Physiol. 278: R1352-R1361. McCormick, S.D., Shrimpton, J.M., Moriyama, S., and Bjornsson, B.Th. 2002. Effects of an advanced temperature cycle on smolt development and endocrinology indicate that temperature is not a zeitgeber for smolting in Atlantic salmon. J. Exp. Biol. 205: 3553-3560. Menard, S. 2001. Applied Logistic Regression Analysis. Sage University Papers Series on Quantitative Applications in the Social Sciences, 07-106. Thousand Oaks, CA: Sage. Muir, W.D., Zaugg, W.S., Giorgi, A.E., and McCutcheon, S. 1994. Accelerating smolt development and downstream movement in yearling chinook salmon with advanced photoperiod and increased temperature. Aquaculture 123: 387-399. Nielsen, C , Aarestrup, K., Norum, U., and Madsen, S.S. 2004. Future migratory behaviour predicted from premigratory levels of gill Na /K -ATPase activity in individual wild brown trout (Salmo trutta). J. Exp. Biol. 207: 527-533. 94 Nielsen, S.E., Johnson, C.J., Heard, D.C., and Boyse, M.S. 2005. Can models of presenceabsence be used to scale abundance? Two case studies considering extremes in life history. Ecography 28(2): 197-208. Paragamian, V.L., and Kruse, G. 2001. Kootenai River white sturgeon migration behavior and a predictive model. N. Am. J. Fish Manage. 21: 10-21. Peake, S., and McKinley, R.S. 1998. A re-evaluation of the swimming performance of juvenile salmonids relative to downstream migration. Can. J. Fish. Aquat Sci. 55: 682-687. Pearce, J., and Ferrier, S. 2000. Evaluating the predictive performance of habitat models developed using logistic regression. Ecol. Model. 133: 225-245. Raymond, H.L. 1979. Effects of dams and impoundments on migrations of juvenile chinook salmon and steelhead from the Snake River, 1966 to 1975. Trans. Am. Fish. Soc. 108: 505-529. Roper, B.B., and Scarnecchia D.L. 1999. Emigration of age-0 salmon (Oncorhynchus tshawytscha) smolts from the upper South Umpqua River basin, Oregon, U.S.A. Can. J. Fish. Aquat. Sci. 56: 939-946. Saunders, R.L., Specker, J.L., and Komourdjian, M.P. 1989. Effects of photoperiod on growth and smolting in juvenile Atlantic salmon {Salmo salar). Aquaculture 82: 101-117. Shrimpton, J.M., Devlin, R.H., McLean, E., Byatt, J.C., Donaldson, E.M., and Randall, D.J. 1995. Increases in gill cytosolic corticosteroid receptor abundance and saltwater tolerance in juvenile coho salmon {Oncorhynchus kisutch) treated with growth hormone and placental lactogen. Gen. Comp. Endocrinol. 98: 1-15. Shrimpton, J.M, and McCormick, S.D. 1998. Seasonal differences in plasma Cortisol and gill corticosteroid receptors in upper and lower mode juvenile Atlantic salmon. Aquaculture 168: 205-219. Shrimpton, J.M., Bjornsson, B.Th., and McCormick, S.D. 2000. Can Atlantic salmon smolt twice? Endocrine and biochemical changes during smolting. Can. J. Fish. Aquat. Sci. 57: 1969-1976. 95 Shrimpton, J.M, and McCormick, S.D. 2003. Environmental and endocrine control of gill corticosteroid receptor number and affinity in Atlantic salmon (Salmo salar) during smolting. Aquaculture 222: 83-99. Sigholt, T., Staurnes, M., Jakobsen, H.J., and Asgard, T. 1995. Effects of continuous light and short-day photoperiod on smolting, sea-water survival, and growth in Atlantic salmon {Salmo salar). Aquaculture 130: 373-388. Smith, S.G., Muir, W.D., Hockersmith, E.E., Zabel, R.W., Graves, R.J., Ross, C.V., Connor, W.P., and Arnsberg, B.D. 2003. Influence of river conditions on the survival and travel time of Snake River subyearling fall chinook salmon. N. Am. J. Fish Manage. 23: 939-961. Stefansson, S.O., Bjornsson, B. Th., Hansen, T., Haux, C , Taranger, G.L., and Saunders, R.L. 1991. Growth, parr-smolt transformation and changes in growth hormone of Atlantic salmon {Salmo salar) reared under different photoperiods. Can. J. Fish. Aquat. Sci.. 48: 2100-2108. Stefansson, S.O., Nilsen, T.O., Ebbesson, L.O.E., Wargelius, A, Madsen, S.S., Bjornsson, B.Th., McCormick, S.D. 2007. Molecular mechanisms of continuous light inhibition of Atlantic salmon parr-smolt transformation. Aquaculture 273(2-3): 235-245. Swets, J.A. 1988. Measuring the accuracy of diagnostic systems. Science 240: 1285-1293. Taylor, W. 2000a. Change-Point Analysis: A powerful new tool for detecting changes. WEB: http://www.variation.com/cpa/tech/changepoint.html. Taylor, W. 2000b. Change-Point Analyzer 2.3 shareware program, Taylor Enterprises, Libertyville, Illinois. Web: http://www.variations.com/cpa Thorarensen, H., Clarke, W.C., and Farrell, A.P., 1989. Effect of photoperiod and various intensities of night illumination on growth and seawater adaptability of juvenile coho salmon {Oncorhynchus kisutch). Aquaculture 82: 39-49. Thorpe, J.E., and Morgan, R.I.G. 1978. Periodicity in Atlantic salmon, Salmo salar, smolt migration. J. Fish Biol. 12: 541-548. Vuong, Q.H. 1989. Likelihood ratio tests for model selection and non-nested hypotheses. Econometrica 57: 307-333. 96 Whalen, K.G., Parrish, D.L., and McCormick, S.D. 1999. Migration timing of Atlantic salmon smolts relative to environmental and physiological factors. Trans. Am. Fish. Soc. 128: 289-301. Zaugg, W.S., Adams, B.L., McLain, L.R. 1972. Steelhead migration: potential temperature effects as indicated by gill adenosine triphosphatase activities. Science 176: 415416. Zaugg, W.S., and Wagner, H.H. 1973. Gill ATPase activity related to parr-smolt transformation in steelhead trout (Salmo gardneri): influence of photoperiod and temperature. Comp. Biochem. Physiol. 45B: 955-965. Zaugg, W.S. and McLain, L.R. 1976. Influence of water temperature on gill sodium, potassium-stimulated ATPase activity in juvenile coho salmon (Oncorhynchus kisutch). Comp Bioc Physiol A 54: 419-421. Zaugg, W.S., Prentice, E.F., and Waknitz. F.W. 1985. Importance of river migration to the development of seawater tolerance in Columbia River anadromous salmonids. Aquaculture 51: 33-47. Zydlewski, G.B., Haro, A., and McCormick, S.D. 2005. Evidence for cumulative temperature as an initiating and terminating factor in downstream migratory behavior of Atlantic Salmon {Salmo salaf) smolts. Can. J. Fish. Aquat. Sci. 62: 68-78. 97 so oo L/i —' L/i O L/i 00 tyi 00 o b (_/i O P- o b » e © O O © © b O O O O O O O O O O o b 3 b o o o o o o o o o o ^ b © © ^— o o o o ^ o k ) i— -J P vJil Ol M OUlO Ul W s Ul l O O t O s l i - t O U U l M b o ^ u u i M i o u w o s u i o o w o o u i i o o o o o o o o o o o o o to © b b b b oo tyi o o o o *- © 3 L J O os 2 -P* O 0 0 * . o L> © U l W < i O l W U l O \ ^ O S 0 0 U M U s l o o to b t o o o o o o o o o o o g i>* •** i ** \s\ -^*\ - 1 - * "*i^ OS U) SO SO OO OO OO -o oo u> -o O © t O U i S 0 O s U J U > 0 S 0 0 0 S U ) S 0 O 3 O P o 8 o b H- o o o P B a. a. B • ~ OS " • to ^ <=> OS *^ o ^ p to to B & SO to S ui OS W SO SO Ln oo SO © OS S= P- © •„ B a, U) O (^ M oo 1-1 -J UJ bs ~ °^ © £ t-o _ S ,= ^- ^ <-" £ P p U * to SO sn £ y 5; g to o> £ bs ^ ^_. OS p -J o § ^ >-> g —«r -J so • K; J2 to M w P ^- P- Ui °^ ^ ~P u, ^O o -J B ^ °* 5 3 a. o. o k> 3 o to ,3 OS W v ) Oi -J -p* -J oo o o 4^ to to 00 o o Cu °* °^ a. 5 o Po -o 3 t O r CD ^ S J i U i U i M ' o O . f t M 0\ » « 0O 4 i U> © b O H » 2 o o o o o o o o o o o o k > o o b o o o o o o o b > g -P^ •£> o o o o to © © b "0 r H W o «< c s 3 aw (re w - y. a. CO to, cr o O to re 3 •a re s a Appendix 2. 3-Factor Analysis of Variance (ANOVA) tables (2005 and 2006 data combined). Factors = sampling event (7 levels), temperature (2 levels), and flow (2 levels). 2a. Fish Length Source Sampling Event (E) Temperature (T) Flow (F) E+T E+F T+F E+T+F Error Sum-of-Squares 58473.46 3233.82 446.31 2267.04 704.47 411.48 740.76 40965.80 df 6 1 1 6 6 1 6 445 Mean-Square 9745.58 3233.82 446.31 377.84 117.41 411.48 123.46 92.06 F-ratio 105.86 35.13 4.85 4.10 1.28 4.47 1.34 P < 0.001 < 0.001 0.028 < 0.001 0.267 0.035 0.237 Sum-of-Squares 20216.69 1437.66 109.73 1814.56 289.71 179.56 381.42 13105.58 df 6 1 1 6 6 1 6 445 Mean-Square 3369.45 1437.66 109.73 302.43 48.28 179.56 63.57 29.45 F-ratio 114.41 48.82 3.73 10.27 1.64 6.10 2.16 P < 0.001 O.001 0.054 < 0.001 0.135 0.014 0.046 df 6 1 1 6 6 1 6 445 Mean-Square 0.612 0.076 0.004 0.026 0.021 0.062 0.015 0.009 F-ratio 65.475 8.091 0.453 2.776 2.208 6.680 1.600 P < 0.001 0.005 0.501 0.012 0.041 0.010 0.146 2b. Fish Weight Source Sampling Event (E) Temperature (T) Flow (F) E+T E+F T+F E+T+F Error 2c. Fish Condition Factor Source Sampling Event (E) Temperature • ( T ) Flow (F) E+T E+F T+F E +T+F Error Sum-of-Squares 3.673 0.076 0.004 0.156 0.124 0.062 0.090 4.160 99 2d. Gill Na+,K+-ATPase Activity Source Sampling Event (E) Temperature (T) Flow (F) E+T E+F T+F E+T+F Error Sum-of-Squares 55.380 3.943 0.006 9.290 2.442 0.968 0.590 62.752 df 6 1 1 6 6 1 6 426 Mean-Square 9.230 3.943 0.006 1.548 0.407 0.968 0.098 0.147 F-ratio 62.659 26.770 0.043 10.511 2.762 6.572 0.668 P < 0.001 < 0.001 0.837 < 0.001 0.012 0.011 0.676 df 6 1 1 6 6 1 6 434 Mean-Square 21.117 13.189 11.243 2.736 2.513 0.436 1.600 0.569 F-ratio 37.145 23.201 19.777 4.812 4.421 0.767 2.815 P < 0.001 < 0.001 < 0.001 < 0.001 < 0.001 0.382 0.011 2e. Plasma Cortisol Concentration Source Sampling Event (E) Temperature (T) Flow (F) E+T E+F T+F E+T+F Error Sum-of-Squares 126.700 13.190 11.243 16.413 15.079 0.436 9.600 246.724 100