Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An examination of differential survival in downstream migrating coho salmon smolts Sawada, Joel O. 1993

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

Item Metadata


831-ubc_1993_spring_sawada_joel.pdf [ 3.07MB ]
JSON: 831-1.0074802.json
JSON-LD: 831-1.0074802-ld.json
RDF/XML (Pretty): 831-1.0074802-rdf.xml
RDF/JSON: 831-1.0074802-rdf.json
Turtle: 831-1.0074802-turtle.txt
N-Triples: 831-1.0074802-rdf-ntriples.txt
Original Record: 831-1.0074802-source.json
Full Text

Full Text

AN EXAMINATION OF DIFFERENTIAL SURVIVAL IN DOWNSTREAM MIGRATING COHO SALMON SMOLTS by JOEL OSAMU SAWADA B.Sc., The University of British Columbia, 1987 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF ZOOLOGY  We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA January 1993 © Joel 0. Sawada, 1993  In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission.  (Signature)  Department of  200LACIf  .  The University of British Columbia Vancouver, Canada  Date  DE-6 (2/88)  JAI) .^I 0'9 9?  ii ABSTRACT  The smolt stage of the life history of Pacific salmon (Onchorhynchus sp.) has received little attention relative  to the other life history stages. I examined mortality during the downstream migration of coho salmon (0. kisutch) smolts by trapping and marking smolts high in the watershed and recapturing the fish near the ocean. Four study sites were used, two in the upper watershed and two in the lower watershed. At both distances, lake - type and stream - type habitats were sampled. Factors which were found to influence survival included: 1) distance of migration, lower survival with increased migration distance; 2) juvenile rearing habitat, stream - type fish had higher survival; and 3) the timing of migration, fish migrating early in the season survived better. A fourth factor examined, the length of the fish at migration, did not influence survival. Although length did not affect survival, it was found that the center of the length distribution often had higher survival than the tails of the distribution, regardless of the absolute length. I performed two separate experiments to assess the effects of handling on fish survival. These experiments did not indicate my handling the fish affected survival. This thesis establishes differential survival occurs in salmon smolts, however the mechanisms causing differential survival are not known.  iii  CONTENTS  ABSTRACT ^  ii  LIST OF TABLES ^  iv  LIST OF FIGURES ^ ACKNOWLEDGEMENTS ^  ix  CHAPTER 1: GENERAL INTRODUCTION ^  1  Study area ^  5  Study sites ^  9  CHAPTER 2: DATA COLLECTION AND EXPLORATORY DATA ANALYSIS ^  10  Introduction ^  10  Materials and methods ^  11  Results ^  21  Discussion ^  63  Conclusion ^  70  CHAPTER 3: STATISTICAL ANALYSES ^  72  Introduction ^  72  Materials and methods ^  72  Results ^  77  Discussion ^  87  Conclusion ^  93  CHAPTER 4: GENERAL CONCLUSIONS ^  94  REFERENCES ^  96  iv  LIST OF TABLE  TABLE 2.1 Day of capture, mark numbers, numbers captured and recaptured, and percent survival for the four upstream sites in 1991 ^ 74  LIST OF FIGURES  FIGURE 1.  Map of study area ^  FIGURE 2.  Possible quantile-quantile plot outcomes ....18  FIGURE 3.  Boxplots of smolt lengths for 1991  ^  24  FIGURE 4.  Boxplots of smolt lengths for 1990  ^  25  FIGURE 5.  Density plots of lengths for  6  Cub Creek 1991 ^ FIGURE 6.  Density plots of lengths for O'Connor Lake 1991  FIGURE 7.  ^  ^  ^  ^  ^  37  Density plots of lengths for Long Lake 1990  FIGURE 13.  36  Density plots of lengths for Misty Inlet Creek 1990  FIGURE 12.  35  Density plots of lengths for O'Connor Lake 1990  FIGURE 11.  29  Density plots of lengths for Cub Creek 1990  FIGURE 10.  31  Density plots of lengths for Long Lake 1991 ^  FIGURE 9.  30  Density plots of lengths for Misty Inlet Creek 1991  FIGURE 8.  29  ^  38  Quantile-quantile^(Q-Q) plots of capture and recapture length distributions for Cub Creek 1991  ^  42  vi  FIGURE 14.  Quantile-quantile ^(4 Q) plots of capture -  and recapture length distributions for O'Connor Lake 1991 FIGURE 15.  ^  43  Quantile-quantile^(Q Q) plots of capture -  and recapture length distributions for Misty Inlet Creek 1991 ^ FIGURE 16.  44  Quantile-quantile^(Q-Q) plots of capture  and recapture length distributions for Long Lake 1991 FIGURE 17.  ^  45  Quantile-quantile ^(Q 4) plots of capture -  and recapture length distributions for Cub Creek 1990 FIGURE 18.  ^  48  Quantile-quantile^(Q-Q) plots of capture and recapture length distributions for O'Connor Lake 1990  FIGURE 19.  ^  49  Quantile-quantile^(Q Q) plots of capture -  and recapture length distributions for Misty Inlet Creek 1990 ^ FIGURE 20.  50  Quantile-quantile^(Q-Q) plots of capture and recapture length distributions for Long Lake 1990  FIGURE 21.  ^  51  Scatterplot of time taken to migrate for each marking period Cub Creek 1991  FIGURE 22.  ^  54  Scatterplot of time taken to migrate for each marking period O'Connor Lake 1991  ^  55  vii  FIGURE 23.  Scatterplot of time taken to  migrate for each marking period Misty Inlet Creek 1991 FIGURE 24.  ^  56  Scatterplot of time taken to  migrate for each marking period Long Lake 1991 ^ FIGURE 25.  57  Scatterplot of time taken to  migrate for each marking period Cub Creek 1990 ^ FIGURE 26.  59  Scatterplot of time taken to  migrate for each marking period O'Connor Lake 1990 ^ FIGURE 27.  60  Scatterplot of time taken to  migrate for each marking period Misty Inlet Creek 1990 FIGURE 28.  ^  61  Scatterplot of time taken to  migrate for each marking period Long Lake 1990 ^ FIGURE 29.  62  Multiple regression response surface of length and timing of migration vs survival  Cub FIGURE 30.  Creek  1991  79  Multiple regression response surface of length and timing of migration vs survival  O'Connor Lake 1991 FIGURE 31.  80  Multiple regression response surface of length and timing of migration vs survival Misty Inlet Creek 1991 ^  81  viii  FIGURE 32.  Multiple regression response surface of  length and timing of migration vs survival Long Lake 1991 ^ FIGURE 33.  Linear least squares regression of  habitat vs survival ^ FIGURE 34.  82  86  Functional relationship between smolt density and percentage mortality by gulls ^ 92  ix ACKNOWLEDGEMENTS  Many people assisted with this project and contributed to its success. I would like to thank my supervisor, Dr. Carl Walters, for suggesting, obtaining funding, and providing support for the project. Bruce Ward and Pat Slaney of the Ministry of Environment and their Keogh River fence staff allowed me to use their site and provided assistance and support during various stages of the project. My research committee, especially Dr. J.D. McPhail, provided valuable comments. The ecology graduate students at UBC provided support throughout the project. I would especially like to thank Dana Atagi, Joe DeGisi, Gordon Haas, Leonardo Huato, and Dick Repasky. Alistair Blachford of the Biological Computing Unit at UBC provided invaluable assistance with data analyses and presentation. Funding was provided by a NSERC/DFO subvention grant awarded to Dr. Walters, Dr. Walters NSERC operating grant, a UBC teaching assistantship, and a UBC Fisheries Centre bursary.  1  CHAPTER 1: GENERAL INTRODUCTION  An important ecological question is which factors influence survival, or mortality, of individuals within a population? This question has been the basis of a myriad of studies in population, community, and evolutionary ecology. With a commercially exploited species, this question has an economic significance as well. Some believe a better understanding of the causes of mortality may lead to an ability to increase production and exploitation. The traditional view of Pacific salmon is there are two major periods of mortality, both involving transitions between environments. The first period is when the fish emerge from the gravel. Apparently, 40-60% of emerging fry die within two months of leaving the gravel (Hartman and Scrivener 1986). The second period of high mortality coincides with the entry of smolts into the ocean (Hunter 1959, Parker 1968, Healey 1982, Murphy et al. 1988). Generally, predation by piscivorous fishes (Mace 1983), aquatic birds (Wood 1987, Mace 1983, Shapavalov and Taft 1954) and mammals (Heggenes and Borgstrom 1988, Ruggles 1980) is thought to be the cause in both mortality periods. With anadromous fishes, the extent of mortality attributable to each life stage is rarely quantified. Mortality of parr has been examined by Wood (1987 b), and by various authors who back-calculated from smolt numbers  2  to obtain parr numbers using an overall fresh water survival rate. (Shapavalov and Taft 1954, Holtby and Hartman 1982). From my literature search, it appears the majority of research on salmonid smolts has focused on the physiology, not the ecology of smolts (Folmar and Dickoff 1980 and references therein). Some research has been done on Pacific salmon hatchery systems (Bilton et al. 1982, Wood 1987); even less work has been performed on wild smolts. Using number of publications as the criterion, the most recent research on wild coho (Onchorhynchus kisutch) smolts originates from two foci. First is the Carnation Creek study examining the effects of logging on coho and chum (Onchorhynchus keta) salmon populations (McMahnon and Holtby 1992 and references therein, Hartman 1982, Hotlby and Hartman 1982). Second is the group of studies examining the impact of impoundments on the Columbia River salmon stocks (Rieman and Beamesderfer 1991, Beamesderfer et al. 1990.) Other studies indicated that survival from smolt to adult can be influenced by factors such as smolt size or smolt migration timing (Healey 1982, Parker 1962, Bilton et al. 1982). The survival of downstream migrating smolts has been examined in Atlantic salmon (Salmo salar) (Bley 1987, Ruggles 1980, Elson 1962, Larsson 1985 and references therein), and sockeye salmon (0. nerka) (Ruggerone and Rogers 1984). The consensus of these reports is that  3  considerable mortality can occur during the smolt migration. This mortality is due to increased vulnerability to predation caused by morphological, behavioral, physiological, and habitat changes associated with smolting and migration. Cameron (1958) estimated 60% mortality for pink salmon (0. gorbuscha) fry migrating to the ocean over a distance of two miles. This mortality varied with the distance the fry had to migrate, and increased during the week by week progression of the run; however, pink salmon migrate as fry whereas coho migrate at a larger size as smolts. A coho study using similar methodology, but different objectives, was performed by Thedinga and Koski (1984). This study did not examine survival but focused on the variability over five years in production, size, and age structure of an Alaskan coho population. Apparently, this thesis is the first study to focus on the freshwater survival of coho smolts during migration. The dearth of research on smolts in their freshwater environment is not surprising. The smolt stage is relatively short (up to 60 days in duration) and facilities necessary to address mortality are not common. The common method of examining mortality during smolt migration involves the use of an enumeration fence as a method of recapture. Using such a facility, Irvine and Ward (1989) documented patterns of changing smolt size through the duration of a smolt run: larger fish migrated earlier in the season. This pattern was consistent over  4  nine years. They also documented a relationship between smolt size and juvenile habitat: lake-reared smolts were larger than-stream-reared smolts of the same age. They did not investigate the effects of these size differences on survival. With a recapture facility and proper logistic support, mortality can be examined by capturing and marking fish at upstream locations and recapturing surviving fish before they enter the ocean. This was the method used in my thesis research. The purpose of my study was to examine survival during the downstream migration of coho smolts, and to determine if survival was influenced by smolt size, migration timing, migration distance, and juvenile habitat.  Hatchery studies suggest that survival from smolt to adult can be increased by manipulating smolt size and migration timing (Hilton et al. 1982). It is not known where in the life history of an individual this effect occurs, but one possibility is the smolt migration stage. If substantial mortality occurs during downstream migration, survival could be related to distance travelled, and therefore time taken, to migrate. Coho smolts from lakes differ in morphology, behavior (Swain and Holtby 1989) and, in the Keogh, in size compared to river fish. I wanted to see if these differences affect survival. The experimental design was to monitor survival, over  5  time, of marked smolts from four rearing locations in a single river.^To provide estimates of intra-annual and inter-annual variations in survival, different marks were applied to smolts migrating at different times within each rearing location and year  STUDY AREA:  The Keogh River is located at northeastern end of Vancouver Island near the city of Port Hardy (127 25 W by 50 35 N) and drains into Queen Charlotte Strait (figure 1). The Keogh is a third order coastal stream approximately 32 km long and drains 130 km 2 of predominantly coniferous watershed (Johnston et al 1990). Thirty-five percent of the watershed has been logged in the past 30 years. The headwater areas above km 29 have been logged to the river, lower sections have buffer strips or have not yet been logged (Johnston et al. 1990). The riparian vegetation in logged areas is red alder (Alnus rubra), salal (Gaultheria shallon), and willow (Salix spp.) The mature forests are composed predominantly of Western hemlock (Tsuga heterophylla), red cedar (Thuja plicata), and sitka spruce (Picea sitchensis). The river has 19 tributaries, drains 6 lakes and displays a highly variable discharge (0.1-225 m 3 s -1 ).  6  Figure 1. Location of the five fish counting fences in the Keogh River watershed. Inset shows the location of the Keogh River watershed' on northern Vancouver Island  7  Main Keagn 7 River Fence -  CUErN CH.4.-7 1 C77=  Mie:y In Ler C:eek Fence  Lang Lake Cutlet Fence  Cut, Creek Fence  Muir Lzke limm=71m== 0 1 2 3. 4 it rn  O'Connor Loke Outlet Fence  Keogh Lae  11.  8  Salmonid research on the Keogh River started in 1975 when a fence was installed for steelhead (0. mykiss) enumeration. Steelhead adult and smolt and coho smolt enumeration continues to the present. Keogh annual salmon escapements are highly variable. In recent years, coho numbers have ranged from 200 to 2000 (Swales et al. 1988). The subsequent number of coho smolts produced per year has ranged from 40 000 to 105 000 (Irvine and Ward 1989).  The fish fauna of the Keogh is similar to many coastal salmon producing streams. In addition to coho, the river supports populations of pink salmon, chum salmon, kokanee salmon (0. nerka), anadromous steelhead trout and resident rainbow trout (0. mykiss), anadromous and resident cutthroat trout (0. clarki), anadromous and resident Dolly Varden char (Salvelinus malma), coast range sculpin (Cottus aleuticus), prickly sculpin (C. asper), three-spine stickleback (Gasterosteus aculeatus), and Pacific lamprey (Lampetra tridentata). Terrestrial vertebrates with the potential to eat coho smolts includes: mink (Mustela vison), river otters (Lutra canadensis), and martin (Martes americana). The avian fauna includes mergansers (Mergus merganser), great blue herons (Ardea herdias), and kingfishers (Megaceryle alcyon).  9  STUDY SITES:  Four sites were chosen for this study. The outlets of O'Connor and Long Lake provided lake-type smolts; Cub Creek and Misty Lake inlet provided stream-type fish. Each habitat type is represented in the upper (Cub and O'Connor, >20 km from the ocean) and lower watershed (Misty and Long, <13 km from the ocean). Long Lake outlet and Misty Inlet streams are small tributaries which enter the main stem Keogh River 11.73 and 12.95 km upstream of the river mouth at an elevation of 90 m (Johnston et al. 1987). There are no lakes above the Misty Inlet sampling site (Irvine and Ward 1989), and juveniles there rear in about 3.5 km of stream. Long Lake is small (approx. 20 ha) and shallow (approx. mean depth 2 m). Due to low summer flows and lack of rearing area, all fish caught at the Long Lake sampling site can be assumed to be from the lake. The outlet stream between the lake and the fence provides very little habitat for juvenile coho salmon. Cub Creek enters the Keogh River about 21.86 km from the stream mouth. There is approximately 3.5 km of stream above the sampling site and there are no lakes on the Cub Creek system. O'Connor Lake is 25.86 km from the river mouth. It has an area of 45 ha and a mean depth of 19 m. Both Cub Creek and O'Connor Lake have an elevation of 185 m.  10  CHAPTER TWO: DATA COLLECTION AND EXPLORATORY DATA ANALYSIS  INTRODUCTION:  This chapter reviews the initial examination of differential survival in coho salmon smolts. In 1990, the collection of field data commenced on April 23 and finished on June 11. In 1991, the field season started on April 20 and ended on July 18. This thesis attempts to address the following questions about the survival of coho smolts: is there a difference in survival attributable to  1. size of the migrating fish? 2. timing of migration? 3. migration distance? 4. juvenile habitat?  Before questions of differential survival are addressed however, it is necessary to establish that differences exist between fish with different migration timings, migration distances, or juvenile habitats. Given my data, the obvious characteristic to focus on is size. This chapter examines whether size differences exist  11  within, and between, systems and between initial capture and subsequent recapture. It also examines factors which may lead to-differential survival. One such factor is the duration of migration. Comparisons will be made within and between systems. If size differences exist, analyses in subsequent chapters will examine if these differences affect survival.  MATERIALS AND METHODS:  Data collection:  During both field seasons fish migrating from O'Connor Lake were trapped using a permanent, horizontal screen smolt trapping facility 50 m downstream of the lake. On Cub Creek and Long Lake, the traps consisted of fyke nets with attached live boxes. On Misty Inlet stream a fence panel trap was used, similar to the system described in Conlin and Tutty (1979). In both years, the traps were operated continuously and fish were sampled as frequently as possible. On average, the traps were emptied every two or three days. At each site, the captured fish were removed from the live box and processed as follows. The fish were anesthetized with dilute 2-phenoxyethanol, measured for length to the nearest millimeter, weighed to the nearest 0.05 gram with a  12  Cahn electrobalance, assessed for condition of smolt on a scale of 1 to 6, marked, and subsampled. The remaining fish were allowed to recover from the anesthetic and then released. The subsampled fish were preserved in formalin and later transferred to isopropyl alcohol. In 1990, 2566 smolts were captured at Cub, 425 at O'Connor, 276 at Misty, and 410 at Long for a total of 3677 smolts. In 1991, 1975 were captured at Cub, 1758 at O'Connor, 1089 at Misty, and 359 at Long for a total of 5181 fish. Fish were marked using a Pan-Jet fish marker (Herbinger et at. 1990, Kelly 1967, Hart and Pitcher 1969) This instrument injects dye subdermally into the fish but because of the small size of coho smolts, dye was only injected into the fins. Two colors were used: the biological stains alcian blue and alcian green. All the fins were used except the dorsal and adipose. The dorsal fin in coho is highly pigmented and because of this the mark was not visible. The adipose was not used because of its small size. Each fin could be marked once, except for the caudal fin which could be marked in three unique positions. Due to the time required to mark each fin, a maximum of three fins were marked on each fish. The number of fins and the number of colors allowed for 672 site and date specific marks. On average, these marks were changed every four days. The fish were recaptured at the main fence located near the ocean (see figure 1). With the exception of fish  13  lost due to fence inefficiency and flooding, all migrating fish are caught at this site. The fish are processed daily by four Ministry of Environment employees who work the main fence and ancillary personnel such as graduate students and field assistants. Since the marked fish account for a small fraction of the total coho smolt run, each fish had to be examined for marks. Early in the first field season, it became evident that a more visible second mark was needed to assist in finding the marked fish. From that time on, the caudal fin was clipped in addition to the PanJet mark. A caudal clip is also used by the main fence crew for their fence efficiency test. Each morning the crew on the main fence sorted and counted the migrating steelhead, coho, Dolly Varden, cutthroat, and cottids. The clipped fish were separated and processed after the main sort. These fish were examined for marks, measured, weighed, and subsampled. After processing and recovery from the anesthetic, the fish were released and allowed to continue their migration. Problems encountered during the first field season, make some of the data suspect. The fences, especially Misty Creek Inlet, were not 100% fish proof and the captured fish may represent a biased sample. Because of the time lag between capture and recapture, a large number of fish were captured and marked before I realized that a caudal clip was necessary. The efficiency of recapture for these fish is low and estimates from these data are  14  suspect. The main fence crew had problems finding the marked fish even after the caudal clip was instituted. Consequently, all data from 1990 are considered suspect. Modifications of methods in 1991 corrected the shortcomings of 1990. For the rest of this thesis, the 1991 data will be presented first and the 1990 data will be used only to support or refute the 1991 assessments.  Exploratory Data Analysis:  Since the data set is large, I first examined the data graphically. The S+ program is ideal for this purpose. It contains functions which do a variety of Exploratory Data Analysis (Tukey 1977) procedures. Density plots (Chambers et al. 1983) display the relative concentration of data points along the measurement scale. To be more precise, the local density at point y is the fraction of data values per unit of measurement that fall in some interval centered at y. The ubiquitous histogram is similar to a density plot but with a different y axis. One could change the histogram into a density plot by dividing the bar heights by the total number of observations and by the interval scale. Even after this change in scale, the histogram would still differ from the density plots because of the discontinuities in the histogram created by going from one interval to another.  15  The density plot deals with the discontinuities by plotting the density of the points within an interval window. The density plots performed by S+ allow the interval window to be a variety of shapes: cosine, gaussian, rectangular, or triangular. By choosing the appropriate shape, the window can smooth the final graphical output. For all my density plots, the gaussian window was used. Empirical quantile-quantile (Q-Q) plots (Chambers et al. 1983) are a method of comparing two distributions. A quantile is similar to the concept of a percentile. The 85th percentile is interpreted to mean that 85 percent of all values fall below this value and 15 percent are above. Similarly, the .85 quantile (Q(.85)) is the point where the fraction .85 of the data lies below and the fraction .15 lies above. The Q-Q plot is constructed by plotting the quantiles of one distribution against the quantiles of another. Post and Evans (1989) demonstrated the technique in their assessment of size-dependent overwinter mortality of yellow perch (Perca flavences). In their study, they had to deal with the possibility of size dependant growth as well as mortality. In this study, comparisons between systems suggest differences in size dependant growth. There is no indication of growth in fish from O'Connor Lake (large body size), but the smaller fish from the other systems may have grown during migration. Because of the small range of sizes within each system, differential growth within a system was not a major concern.  16  If the data sets are expressed as Xi for i = 1 to n and Yj for j = 1 to m, the quantile-quantile plot is a plot of Q y (p) versus Q x (p) for a range of p values. For example, if the median is plotted against the median, the Q-Q plot is Q y (.5) vs. Q x (.5), if the upper quartile is plotted against the upper quartile the Q-Q plot is Q y (.75) vs. Q x (.75), and so on. If the two distributions are identical, the resulting quantile-quantile plot is a straight line (y=x figure 2, graph 1). Any deviation from this line indicates differences between the two distributions. A regression line drawn through the data which lies parallel to the y=x line indicates that the distributions are the same shape but, that one of the distributions has higher quantile values; growth could produce this result (figure 2, graph 5). A line that is not parallel to the y=x line indicates size selective mortality or growth has occurred. A line with slope of < 1 that is above the y=x line could be due to size selective mortality on smaller individuals (figure 2, graph 2). Similarly, a line with slope < 1 falling below the y=x line could indicate size selective mortality of larger individuals (figure 2, graph 3). A line with slope < 1 that intersects the y=x line could indicate size selective mortality on both tails of the distribution; larger and smaller individuals suffer greater mortality than the middle sized individuals (figure 2 graph 4).  17  Figure 2. Eight possible outcomes of the quantile-quantile (Q-Q) plots. Graphs 1-4 depict possible outcomes in the absence of fish growth. Graphs 5-8 represent possible outcomes when growth is present. Graphs 1 and 5 represent the Q-Q plots in the absence of size selective mortality, graphs 2 and 6 are outcomes under size selective mortality of small individuals, graphs 3 and 7 are outcomes under size selective mortality of large individuals, graphs 4 and 8 are outcomes with size selective mortality of small and large individuals.  18  GROWTH absent 1  ^present  ^  /  none  /  /  7 large  V z /  4 small & large  /  /  /  /  /  19  If growth has occurred, the y intercept of the line is increased by the growth increment and interpretation of the results is necessary. The graph of growth and large individuals suffering greater mortality (figure 2 graph 7) is similar to the case where there is no growth but smaller and larger individuals suffer greater mortality (figure 2 graph 4). If growth occurs and smaller and larger individuals have greater mortality (figure 2 graph 8), the graph is qualitatively the same as the graph for no growth and smaller individuals experiencing greater mortality (figure 2 graph 2). To complicate matters further, the increment of growth can lead to ambiguity. A small increment of growth with small and large individuals suffering greater mortality (figure 2, graph 8) can look similar to a large increment of growth with large individuals suffering greater mortality (figure 2, graph 7) . To aid in distinguishing between the possible causes, an examination of the data points, regression line, and density plot distributions is undertaken. The following procedure was used. The first thing looked for was growth. This could be indicated in the density plots by the largest values of the capture and recapture distributions. If the largest fish recaptured is larger than the largest fish captured, growth or measurement error are the two possible causes. If the mode of the recapture distribution is larger than the capture distribution and the difference is  20  of a similar magnitude as the difference between the largest capture and recapture values, growth, not measurement error, is assumed. If the growth occurs in -  some marking periods, it is likely that it occurs in other periods. After the presence or absence of growth is established the density plots must be examined to distinguish between small and large increments of growth. If data points on the QQ plot show no effect in the center of the distribution but higher mortality on the ends of the distribution, the conclusion is size selective mortality for both the largest and smallest size classes. A recapture distribution narrower than the capture distribution also indicates that fish in the tails of the size distribution suffered greater mortality.  Monte Carlo simulations were used to determine the effect of sample size and random mortality on the QQ plots. A random normal length distribution of sample size 200, 100, 50, 20, and 10 fish represented the capture distributions. These distributions were subjected to random mortality of 50% to obtain a recapture distribution. Fifty simulations for each sample size were performed and QQ plots were used to compare the simulated capture and recapture distributions. For each sample size, the fifty QQ plots were examined to determine what qualitative outcomes could be due to random mortality. These plots were used as criteria for the actual results to determine  21  which QQ plots were not due to the null hypothesis of random mortality. Scatterplots were used to present the number of days until recapture for each marking period in each system (figures 13 -20). The day of capture was the x axis and the number of days until recapture was the y axis. The data points were jittered on both axes to distinguish individual data points.  RESULTS:  Boxplots.  The first exploration of the data was to make boxplots of the fish lengths for each system and marking period to display trends over time (figures 3 and 4). An examination of the four series of boxplots from 1991 (figure 3) reveals four clear patterns. First, the recaptured fish were consistently larger than the initial captures. This was most notable in Cub Creek and least notable in O'Connor Lake. Fish from the lake systems, O'Connor and Long, were longer than the fish from the stream systems, Cub and Misty. Second, the duration of the smolt run, as estimated from the number of marking periods, was longest in the upstream systems. Third, in all systems except Misty, the variance in body size decreases as the  22  smolt run progresses and, finally, the size of the fish from O'Connor Lake steadily increases throughout the season. This pattern was not apparent in the other systems.  Boxplots of the 1990 data (figure 4) reveal the median sizes of the fish in all systems were smaller than in 1991, and the difference in size between capture and recapture was not as pronounced. The fish from O'Connor Lake were still the largest, but the fish from Long Lake were no larger than the fish from Misty Creek. The Cub Creek data show a number of outliers greater than 160 mm in length. These larger fish were not evident in 1991.  Density Plots.  To examine the length data further, density plots were graphed for all systems and all marking periods. These plots provide a different view of the distributions than the boxplots, and allow for identification of traits such as bimodality. When captures and recaptures are plotted on the same graph, density plots effectively display the size and shape changes of the distributions from capture to recapture. If changes in the shape of the distribution occur, this indicates size related differential mortality;  23  Figures 3 and 4. Boxplots of length of captured and recaptured smolts from the four study sites in 1991 (figure 2) and 1990 (figure 3). Day of capture is on the x axis. Capture and recapture distributions are adjacent to each other; different marking groups are separated by the vertical dashed lines. If there were no recaptures, only the capture boxplot lies between two vertical dashed lines. The range of the data is represented by the outer horizontal dashes. The center 50% of the data is represented by the solid vertical line. The "+" symbol, often seen as the inner horizontal dashes, represent the 95% confidence interval of the median. The median is represented by a single dot. The smaller points lying outside the range are data outliers.  N .A.  tv tri  26  however, the density plots are not as effective as the boxplots in detecting trends in length over time. The density plots for Cub Creek 1991 are displayed in figure 5. The marking period is identified by the number in the top left corner of each graph. This number represents the day of the year (January 1 being day 1 and December 31 being day 365) of the middle day of each marking period. Graphs for days 151 through 168 represent only small numbers of recaptures. For the remaining graphs, it appears that both the capture and recapture distributions which were initially complex, become more simple over time. Throughout the season the capture distributions were skewed to the right. The capture and recapture distributions were initially different in size and shape but, over time became more similar. ^The recaptures in graph 113 display a large shift in the mode and a reduction in the skew. The remaining graphs maintain the skew in the recapture distributions and show more similarity between the modes of the capture and recapture distributions. The fourteen graphs for O'Connor Lake 1991 (figure 6) show different trends than Cub Creek 1991. Graphs 144, 155 through 171 represent small numbers of recaptures. ^The most noticeable feature of these graphs is the similarity in shape between the capture and recapture distributions. Of equal importance is the absence of the larger and smaller sizes in the recaptures of graphs 120, 124, 141,  27  and 147. Graphs 137 and 148 indicate that for these dates, the larger size classes survived poorly; the smaller size classes did poorly for date 151. -  Misty Creek 1991 data are displayed in figure 7. Of the 11 graphs, the last four represent low numbers of recaptures. Misty Creek behaved similarly to Cub Creek in two noticeable ways. The size and shape differences between capture and recapture were large for the early capture sessions; this difference decreased as the season progresses. The smaller size classes were not present in most of the recapture distributions. Long Lake (figure 8) displayed the shortest time duration of migration with only six marking periods in 1991 for a total duration of 23 days. The graphs of Long Lake were similar to O'Connor Lake in that the distributions were generally symmetrical. Unlike O'Connor Lake, there was an increase in size of recaptures in the first three marking periods. The final three marking periods display similar modes in the capture and recapture distribution.  The density plots for 1990 are displayed in figures 9 through 12. To eliminate suspect data, only the marking periods which included caudal clips are presented. For Cub Creek 1990 (figure 9), the first four marking periods are not displayed due to their lack of a caudal clip. In graphs 127 and 132, the capture and recapture distributions are very similar; there is no indication of  28  Figures 5 — 8. Density plots of the capture and recapture distributions for the four study sites: Cub Creek (figure 5); O'Connor Lake (figure 6); Misty Inlet Creek (figure 7); and Long Lake (figure 8) in 1991. The distributions of captured fish are represented with open circles, the recaptures are depicted with closed dots.^The number in the upper left corner of each graph is the middle day in each capture period.  Cub Creek 1991  Length (mm)  o captures . recaptures  O'Connor Lake 1991 0  124  0  130  137  140  a  0  0  0 100 120 140 160 180 200  0  141  A  7  000000004 ^4rte••0000 0o 100 120 140 160 180 200  8  8  0  0  g  0  100 120 140 160 180 200  0  0  0  0 %IC  100 120 140 160 180 200  0  148  0  00  (wow,*6•0^lNi•ocion000  100 120 140 160 180 200  0  0  •  U)  Q) 0  0  .0t 0  4) •  0 0  8  0  .0  6  0000000000000J 11‘0,3  155  0  0  0 100 120 140 160 180 200  0  157  •• •  0 100 120 140 160 180 200  0  168  " :sp:  100 120 140 160 180 200  171  03 00  0  0  0  00 00  0  00 0 0 00  0 0  0  2 100 120 140 160 180 200  0  tp-. .0  100 120 140 160 180 200  0  8  0  100 120 140 160 180 200  I  '  100 120 140 160 180 200  Length (mm)  100 120 140 160 180 200  o captures . recaptures  100^120 140 160 I80 200  Misty Creek 1991 An  9  120  0  0  8 0  • o.dooe  °  0  60^ao^100^120^140  8  C  • oC4:1366italillic°  60^ao^100^120^140  60^ao^100^120^140  60^ao^100^120^140  147  . (1)  a  0  0  9  9  0  0  0  0 0  60^60^100^120^140  A  kAk  c.! 0  0  60^ao^100^120^140  A  •  O 0^• 0^.n; • 0^• 0 Air •  60  BO^100^120^140  60  BO^100^120^140  60^80^100^120^140  vi  148 0 0  0  •■•  A 0 0  Av  157^a •• •• •• ••  C  0  0  • ••  0  06^.6  0.  •  0 0  0  60^ao^100^120^140  60^80^100^120^140  Length (mm)  o captures recaptures  LJ  33  growth or size selective survival. More recaptures are represented in graphs 138 and 142; graphs 146 through 161 represent only small numbers of recaptures. Of the four graphs for O'Connor Lake 1990 (figure 10), only graph 127 has sufficient numbers of recaptures to allow interpretation. This graph shows the distribution of the recaptures having a considerably smaller range than the capture distribution. Both small (< 125 mm) and large (> 175 mm) fish were present in the captures but not the recaptures. Misty Creek 1990 is represented by the six graphs in figure 11. Graphs 133, 148, and 156 represent small numbers of recaptures. The distribution of the initial captures in graph 127 is initially multimodal; the distribution of the recaptures is bimodal. On this date, the larger fish did poorly; those with length >118 did not appear in the recaptures. Conversely, graph 137 shows a recapture distribution with a lack of the smaller sizes and better survival for larger fish. Graph 142 shows the largest and smallest fish doing poorly. Long Lake 1990 is represented by the four graphs in figure 12. Of these four, graph 137 has no recaptures. The remaining three panels show a remarkable similarity between the capture and recapture distributions.  34  Figures 9- — 12. Density plots of the capture and recapture distributions for the four study sites: Cub Creek (figure 9); O'Connor Lake (figure 10); Misty Inlet Creek (figure 11); and Long Lake (figure 12) in 1990. The distributions of captured fish are represented with open circles, the recaptures are depicted with closed dots.^The number in the upper left corner of each graph is the middle day in each capture period.  Cub Creek 1990 0  0 0  80  127  8  132  0  80  80 30 0  0  °Iet455%•••••••m~......sorms.  00 04 • 100  3  0  80  8  0  0  ^  150  0  0 100  200  200  150  200  150  100 0  0  0  0  146  0  74  30  3  0  0 0 150  100  200  0  100  30 0  200  150  0  100  ^  150  ^  200  0 0  80  0  0  157 •  80 0 0  8 0  0 0 100  ^  150  ^  0  200  ^  8  ••  30  CI •  o^° • Afieja t  p  100  ^  161  0  8  150  ^  0  200  ^  103^150  o captures  Length (mm)  ^. recaptures  ^  '400  ^  O'Connor Lake 1990  132  0  CI  O  ilt ^• , 1i^esik ALL 810.140stitemt lllll Oho  ^11  ;^to"tili Ito t "lt it  0  100^150^200^250^300  100^150^200^250^300  1 137 0  ci  0  00  •• •• •• •• ••  I  0  100^150^200^250^300  ^  100^150^200  ^  250  ^  o captures  Length (mm)  . recaptures  300  Misty Creek 1990 137  O  0  8  • ••  0  0  0  cia•  oo • ocP/  •  ••  •  BO  100  120  140  160  60  BO  100  120  140  160  60  0  130  100  120  140  160  156  0  0  0  60^80^100  120  140  Length (mm)  160  60^80  .0  •  0  •  0 o  °  100  120^140^160  o captures . recaptures  Long Lake 1990  0  0  o 0  6  6  Id  1  .4-  0  q  0  80  100  120  140  80  100  120  140  o_ 6 40 6 ,  6  q 0 80  ^  100  ^  120  ^  140  q 0 ^  80  ^  100  ^  120  ^  o captures  Length (mm)  ^. recaptures  140  39  Quantile—Quantile Plots 1991.  The density plot of Cub Creek 1991 displayed the most dramatic difference between capture and recapture distributions. QQ plots (figure 13), of days 114, 140, 165, and 168 could be due to random mortality. The graphs of days 113, 120, 124, and 128 indicate size selective mortality has occurred. The QQ lines have positive deviations and have a slope < 1. This outcome could be due to either no growth with larger fish surviving better than the smaller fish (figure 2, graph 2), or growth occurring and the larger and smaller size classes surviving poorly (figure 2 graph 8). Panels representing days 132 and 137 display the regression line with slope < 1 intersecting the 1:1 line. This outcome could be due to: (1) no growth; (2) a small increment of growth with smaller and larger fish surviving poorly (figure 2, graph 4); or (3) a large increment of growth with larger fish surviving poorly (figure 2, graph 7). Panels 147 and 151 are of use since they show growth and no size selective mortality. This indicates growth has occurred in this system. An examination of the data points of the regression and the density plots indicates the fish from days 132 and 137 were growing and smaller and larger fish were surviving poorly. The density plots of O'Connor Lake 1991 displayed the smallest difference between capture and recapture. The QQ plots (figure 14) display the differences which exist. The  40  panels representing days 120, 130, 137, 140, 144, and 155 through 171 could be due to random mortality. The regression line on panels representing days 124, 141, and 147 through 151 intersect around the middle of the y=x line.^Once again, this could be due to no growth and greater mortality of both the smaller and larger size classes or growth and greater mortality in the larger size classes.^The box and density plots indicate little growth has occurred. One can conclude the QQ plots represent greater mortality for the larger and smaller size classes. Misty Creek 1991 (figure 15) has six panels, the days 124, 137, and 147 through 168, which the QQ plots could be due to random mortality. The QQ plot of day 116 displays a large increment of growth. The regression line in Panel 120 appears to be caused by slight growth and higher mortality for smaller and larger size classes. Since growth is evident in panel 116, panels 129, 132, and 138 indicate size selective mortality against smaller and larger individuals.^Long Lake 1991 (figure 16) had only six marking periods. Due to the small number of captures in each of the marking periods, the results of all the QQ plots for this system in 1991 could be due to random mortality.  41  Figures 13 - 16. Quantile - quantile (Q-Q) plots of the four systems: Cub Creek (figure 13); O'Connor Lake (figure 14); Misty Inlet Creek (figure 15); and Long Lake (figure 16); in 1991. The scale of the x and y axes is millimeters, the dashed line represents x=y. The points are the actual Q-Q data points; the solid 'line is a least-squares regression. The number is the upper left corner of each graph is the middle day of each capture period.  Cub Creek 1991 0  0  N  0  0  8  0  8  8  O  DO^100^110^120  80^GO^100^110^120^130  90^100^110^120  U  n  BO^90^100^110^120^130  0  0  Cl. o  0  8  8  te,  a  8  CD  CC  BO  90  103  110  90^100^110^120^130  140^80^90^100^110^120^130^140^80  0  165  00  110  100  0  0  0  8  8  2  2  0  8  BO^90^100^110^120  90  100  110  120^70  Captures  80  00  100  110  70  BO  90  100  110  O'Connor Lake 1991 137  6  8  F.  0  8  8 ..  0 7..  120^140^160^180  140^160^180  F.  140 145 150 155 160 165  120^140^160^180^200  120^140^160^180 a a .-  N  r:  8 8  , 8  F.)  2  0  8  8 140^150^160^170  120^140^160^180  157  140^150^160^170^180  168  130 140 150 160 170 180 100^130^140^150^160^170^180  ii  . te  o m  8  8 ,..  0  0  160^165^170^175^160^180^200^220^240  171  'S. 160^180^200^220^240  Captures  160^180^200^220^240  Misty Creek 1991 N N N  0  8  8  8  8  0  8 -  80^90^100^110^80^90^100^110^120^130  90  100  110^120^80^90^100^110^120^130  N  0 8  8  8  8  a a a BO^90^100^110^120  80^90^100^110^120^130  60  90  100  110^120^130  N  N  8 8 92^94^96^98^100  95^100^105^110  90^100  Captures  110  120  130  85  90  95  100  105  Long Lake 1991 A  F  8 0 0  8  0  105  120  115  10  125  90  (/) 0 0  0  CZ  8  o  (1.)  CC  100  1 0  120  130  0  V.V.  N  8  8  8 100  110  120  130  100  Captures  46  Quantile-Quantile Plots 1990.  The density plots for Cub Creek 1990 which had sufficient numbers of recaptures displayed little difference between capture and recapture. The QQ plots of the data (figure 17) indicate subtle differences exist in some of the marking periods. The outcomes of the QQ plots representing days 142, 146, 148, 152, 157, and 161 could be due to random mortality. Panels 127 and 132 show little difference between distributions. Panel 138 displays a slightly higher mortality for the smaller and larger fish. O'Connor Lake 1990 is represented by only small numbers of recaptures, the QQ plots are displayed in figure 18. Only the panel representing 127 has sufficient numbers of recaptures, the other QQ plots could be a result of random mortality. If no growth has occurred, the QQ plot for day 127 shows higher mortality for both larger and smaller individuals. Misty Creek 1990 (figure 19) had panels representing 6 marking periods. All six QQ plots could be a result of random mortality. Long Lake 1990 (figure 20) had four marking periods, all of which could be a result of random mortality.  47  Figures 17 - 20. Quantile - quantile (Q-Q) plots of the four systems: Cub Creek (figure 17); O'Connor Lake (figure 18); Misty Inlet Creek (figure 19); and Long Lake (figure 20); in 1990. The scale of the x and y axes is millimeters, the dashed line represents x=y. The points are the actual Q-Q data points; the solid line is a least-squares regression. The number is the upper left corner of each graph is the middle day of each capture period.  Cub Creek 1990  8 a a  8 2 80  100  120  140  160  200  180  80  90  100  110  120  90  100  110  120  43  o  2  a o as  a 80  40  100  110  120  90  95  100  106^  80^85^PO^95^100^105^110  ...'  152  .....  .•  ..•  .••  ..'  .•"  0  .••".. 0  8 ... 0 o  BO^100^120^110^160^180^200  80^85^90^95^100^105^110  Captures  90  100  110  120  130  O'Connor Lake 1990  1.1) Li)  0 CD 1..-  0 vt  U)  .;r ...--  1-  120 140 160 180 200 220 240 260  140  160  180  200  137 0 0 ir-  0 CO  140^160^180^200^120^140^160^180^200  Captures  Misty Creek 1990 0 o  0  8  !  2  80  po  100  110  100  130  120  110  110  150  80  100  80  110  120  8 8  li  a  0 o  a 80  ^  90  ^  100  ^  110  ^  85  ^  90  ^  95  ^  100  ^  Captures  105  ^  110  ^  90  ^  95  ^  100  ^  105  Long Lake 1990  U,  0 ..--  0 T.... 1-...  U) co  0 0)  90^100^110^120^  90^95^100^105^110^115  137 to . • - • ... ... ' .  .••.*.•. .•-• .....----"--  ... - .  o)  0 co in oo  90^100^110^120^  90^92^94^96^98^100 102  Captures  52  Scatterplots 1991.  The X-Y scatterplots revealed similarities exist among Cub (figure 21), Misty (figure 23) and Long (figure 24). In all systems except O'Connor (figure 22), the time taken to migrate decreased as the season progressed. In the initial marking period for the three similar systems, the first recapture occurred approximately 20 days after initial capture; the last fish were recaptured approximately 50 days after capture. From this point onward, the number of days until first recapture decreased to approximately 5 days near the end of the migration. The fish from O'Connor Lake displayed a different pattern. In the initial marking period, the first recapture occurred 9 days from capture and the last recapture occurred 21 days after capture. The time until recapture decreased slightly over the smolt run, reaching a low of 4 days for the fish released on day 138. In all systems, there was a decrease in the variability of time spent migrating at the end of the smolt run.  Scatterplots 1990  The scatterplots for 1990 suffer from the fact that only the later marking periods are represented. The first data points for Cub Creek 1990 (figure 17) represent fish from day 125. At this time, the fastest migrating fish  53  Figures 21 — 24. Scatterplots of each system: Cub Creek (figure 21); O'Connor Lake (figure 22); Misty Inlet Creek (figure 23); and Long Lake (figure 24), in 1991 displaying the time taken to migrate from the upstream site to the main fence. Each data point represents a single fish. The data points are jittered on the x and y axes to aid in distinguishing each point. The x axis is the middle day of each marking period.  III  58  Figures 25 - 28. Scatterplots of each system: Cub Creek (figure 25); O'Connor Lake (figure 26); Misty Inlet Creek (figure 27); and Long Lake (figure 28), in 1990 displaying the time taken to migrate from the upstream site to. the main fence. Each data point represents a single fish. The data points are jittered on the x and y axes to aid in distinguishing each point. The x axis is the middle day of each marking period  (I1 L.f..)  Misty Creek 1990 • • • •  •  •  0  •  •  •  •  S.  •  • 130  ^  135  ^  140  ^  145  Day of Capture  ^  150  ^  155  63  took only a few days to be recaptured. The slowest fish took close to 30 days to migrate. The three other marking periods with adequate numbers of recaptures display a decreasing range of values. The decrease is due to the upper values diminishing rather than the lower values increasing. O'Connor Lake 1990 (figure 18) has only one marking period with sufficient recaptures (day 125). Here, the shortest migration duration was only a few days and the longest was approximately 17 days. Misty Creek 1990 (figure 19) and Long Lake 1990 both had three marking periods with sufficient recaptures. The scatterplots do not reveal any trends in the data.  DISCUSSION:  The boxplots revealed that size differences exist both between and within systems. The observed length differences between systems for 1991 were consistent with the 1990 data and other previous findings (Irvine and Ward 1989). This indicates the observed size differences between systems are consistent from year to year. However, the length of smolts within a system are not constant from year to year; for all systems, the fish in 1990 were approximately 10% smaller than the fish in 1991. This is to be expected since the factors which affect growth, such  64  as water temperature, are variable between years. Another notable difference in between system comparisons was that larger fish-originate from the lake type habitat. It has been documented that morphological and behavioral differences exist between lake and stream type fish (Swain and Holtby 1989), but there is no documentation of size differences. Throughout the 1991 smolt run, the observed within system size changes were similar between the two creek systems. There did not appear to be any consistent similarities in size changes in the lake systems or for upstream - downstream comparisons. The 1991 boxplots (figure 3) reveal the median size of fish from Cub Creek and Misty Creek increased for the first three marking periods and then decreased as the season progresses. The size of the fish from Long Lake remained approximately the same for the first there marking periods and then decreased. The median size of fish from O'Connor Lake increased throughout the smolt run. It has been documented that lake-type fish are more homogeneous in size than their stream type conspecifics. (B. Holtby, pers. com ). The pattern displayed by the creek fish may be due to the proportion of 2+ smolts migrating. Irvine and Ward (1989) found that the proportion of 2+ smolts changed throughout the migration; they attributed the trend in mean length to these older, bigger fish. Irvine and Ward (1989) also documented predominantly age 1+  65  fish emigrating from O'Connor Lake (86% age 1+) and Long Lake (91% age 1). In comparison, they found only 49% age 1+ fish from their sample in Misty. It was not documented at what time in the smolt run these samples were taken. These differences in the proportion of 1+ fish could explain why the stream systems behave differently from the lake systems. The fact that the lake systems do not behave similarly when compared to each other may be due to the unusually large fish in O'Connor Lake.  Cub Creek, Misty Creek and Long Lake were the systems where non-overlapping boxplot notches occurred; the boxplot notches represent 95% confidence intervals of the median. Cub Creek's significant differences occur throughout the smolt run while Misty and Long displayed these differences only in the initial marking periods. O'Connor Lake showed little difference in length between capture and recapture. These differences are explained by the time taken to migrate. Examination of the scatterplots reveal the time until recapture changes throughout the smolt run. In every system except O'Connor, the number of days until recapture decreased as the season progresses. This is consistent with the diminishing difference in length between capture and recapture as the season progresses. As the migration time decreases, the fish have less time to grow. These changes in length could be due to two factors, growth or mortality. Early in the season, the fish take  66  longer to migrate and have more time to grow. The opportunity for size selective predation is greater because the fish are vulnerable for a longer period of time. The results from O'Connor Lake were not similar to any of the other systems. The boxplots reveal the size of the fish were significantly larger than all other systems. Unlike the other three systems, the size of the migrating fish increased throughout the smolt run. The scatterplot of the time taken to migrate displays a shorter time than the other three systems. While the first migration period took approximately 20 days for Cub (figure 13), Misty (figure 15) and Long (figure 16), the period was only 10 days for O'Connor (figure 14).^This could be due to the large size of fish from O'Connor. Swimming speed is known to be related to body size; larger fish can swim faster. For O'Connor Lake, the time to migrate decreased to about 5 days at day 140 then increased to 25 days on day 141. After this day, the time to migrate slowly decreases until the last marking period. The boxplots (figure 3) reveal the size of the fish migrating on day 141 was significantly larger than for fish migrating one period earlier. These larger fish migrated more slowly than the smaller fish from the previous marking period. This contradicts the general observation that the larger O'Connor Lake fish migrate more quickly than the smaller fish from the other systems. It is possible that there were two types of fish migrating from O'Connor Lake. Irvine and Ward (1989) found that the  67  fish they described as "early" and "late" had different commercial catch rates and distributions. The early migrating smolts were caught in northern areas and the late migrating smolts were caught in southern areas. However, Irvine and Ward (1989) caught and tagged their fish at the main fence and it is impossible to determine where these smolts originated. Therefore, this information does not provide strong support for the idea of two types of fish in O'Connor. The boxplots for 1990 (figure 4) show little similarity to those of 1991 (figure 3). As was mentioned before, the mean length of the fish were approximately 10% smaller in 1990 than 1991 and there was very little length difference between the captures and recaptures. There were no significant differences between capture and recapture in any of the systems. It appears that size selective mortality, or growth, did not occur to the same degree as in 1991. However, in 1990 we have data from only the later marking periods. Most of the size selective mortality and growth occurred early in 1991. An examination of the graphs of the time taken to migrate reveals that in some cases the time taken in 1990 was shorter; this means the fish were vulnerable for less time and had less time to grow. This shorter migration time may be a factor in the observed differences between years. The QQ plots were used to distinguish between growth and differential mortality. These plots reveal some  68  similarities within and between systems. Without question, size-selective mortality did occur, and predominantly affected the smaller and larger size classes.^The most interesting observation is that fish from the middle of the size distributions had higher survival regardless of what absolute length the middle of the distribution corresponded to. The median size of the recaptures, which we can use as an indication of the center of the distribution, was different between systems at the same time and changed over time within a system (figure 23). This means the fish were not selected with respect to length but with respect to position in the distribution.^A possible mechanism lies in the fact the fish school during their migration. In other schooling systems (juvenile kokanee salmon or clupeid fishes are good examples), the length distribution is extremely narrow. This is usually attributed to predation. If predation is a factor, a fish in a school has a lower probability of mortality if it can avoid being noticed. The best way to avoid notice is to be similar to the other fishes. If this is true, the tails of the distribution will have a higher probability of being killed. ^Another possible mechanism is that fish could segregate into schools of similar size and swimming performance. Since the tails of the distribution have smaller numbers, the schools would be smaller and the probability of mortality for each fish is higher. Higher survival for the center of the distribution is  69  not observed in all cases. In Cub Creek 1991 (figure 13) only growth was observed at marking periods 147 and 151. Misty Creek-1991 (figure 15) displayed only growth in period 116. I am at a loss to explain these observations and can only speculate. The possible effect of small sample size has been accounted for with the Monte Carlo simulations; these results were not due to random mortality. It would appear the mortality pressures were not consistent over time. This could be due to effects like predator switching prey types or mobile predators leaving the system. The results of the QQ plots for 1990 were not as clear as the plots for 1991. For example, Cub Creek 1990 (figure 17) panels 127 and 132 display little growth or mortality. Panel 138 displays either growth and higher mortality for smaller and larger sizes or no growth with smaller individuals suffering higher mortality. Since the marking period one plot earlier shows no growth and the marking period one plot later shows growth, no conclusions can be drawn about what was actually happening. An examination of the density plot (figure 9) suggests that growth and higher mortality for the smaller and larger size classes is the most likely mechanism. The range of recapture values was smaller than those of the captures and the data points for the recaptures were shifted, indicating growth. In this system in 1990, only one of three marking periods displayed higher survival for the middle of the distribution.  70  O'Connor Lake 1990 had only one QQ plot with sufficient recaptures. This plot showed higher survival for the center of the distribution. ^Because of the small sample sizes in Misty Creek and Long Lake 1990, all the QQ plot results could be attributed to random mortality. The data from 1990 do not provide strong support for the 1991 data. Size selective mortality favoring the center of the distribution was not as prevalent. It should be mentioned that the 1990 data exist only from day 127 and many of the samples sizes were small. The majority of the selection for the middle of the distribution occurred early in the year in 1991. Had the 1990 data set been as extensive as the 1991 data set, the results may well have been more similar.  CONCLUSION:  It appears size selective mortality occurs in coho salmon smolts. However a single size is not selected for, nor is a trend favoring smaller or larger fish. Quantile quantile plots reveal the center of the distribution often has higher survival than the tails of the distribution. This outcome occurred for a wide range of lengths within and between systems. A possible explanation for the center of the distribution having higher survival is schooling. Two mechanisms can be proposed. Either the tails of the  71  distribution have higher predation rates because they are more visible and therefore more vulnerable, or segregation occurs with-respect to size. In the case of segregation by size, the fish representing the tails of the distribution would have smaller schools and the vulnerability per individual would be higher.  72  CHAPTER 3: STATISTICAL ANALYSES  INTRODUCTION:  This chapter presents results of statistical examination of factors which may lead to differential freshwater survival of coho salmon smolts. The effects of size and timing on survival were examined using multiple regression analyses. The effect of habitat on survival was examined using a single variable regression. To assess effects of migration distance on survival, fish were transplanted downstream while control fish were released from their upstream sites. In addition to natural mortality factors mentioned above, effects of handling and anesthetic on survival were examined in two separate experiments.  MATERIALS AND METHODS:  Questions 1 and 2: do fish survive differently with respect to size at migration and timing of migration?  For each marking period in each system, survival rates were calculated as the ratio of recapture to capture numbers (see table 2.1). For statistical comparisons, survival estimates were transformed using a square-root  73  Table 2.1. Day of capture, mark numbers, numbers captured and recaptured, and percent survival for the four upstream study sites in 1991.  ^  74  Day^#Captured^Survival Site^Mark #^#Recactured Cub^113^117^30^13^0.43 Creek^114^106^48^29^0.60 120^112^161^67^0.42 124^104^235^83^0.35 128^121^81^42^0.52 132^120^579^202^0.35 137^133^418^147^0.35 ^1 40^138^122^47^0.38 147^127^141^52^0.37 151^140^28^10^0.36 165^143^10^1^0.10 168^167^9^0^0.00 O'Connor 120^101^143^56^0.39 Lake^124^107^136^33^0.24 130^125^35^20^0.57 137^134^430^152^0.35 140^139^221^21^0.09 141^154^26^15^0.58 144^145^50^3^0.06 147^130^162^14^0.09 148^156^206^6^0.03 151^149^113^9^0.08 155^154^126^15^0.12 157^157^21^2^0.09 168^166^13^1^0.08 171^169^6^0^0.00 Misty^116^128^28^9^0.32 Creek^120^109^83^23^0.28 124^110^53^27^0.51 129^124^105^46^0.44 132^116^366^132^0.36 137^131^217^79^0.36 138^135^87^20^0.23 147^122^8^4^0.50 148^148^6^3^0.50 157^155^8^3^0.37 168^164^11^2^0.18 Long^113^103^37^23^0.62 Lake^116^115^25^12^0.48 120^108^77^46^0.60 124^113^45^30^0.67 129^119^63^25^0.40 135^129^49^12^0.24  75  arcsine transformation. Multiple regression was used to estimate the relationship between the transformed survival data and: l} timing of migration; 2) median length at capture; 3) median length 2 at capture; and 4) median length 3 at capture. Polynomial parameters were included in the initial model to determine if survival was related to a power function of length. A stepwise multiple regression procedure was used to determine the best set of parameters and the response surfaces were plotted.  Question 3: do fish survive differently with respect to migration distance?  In 1991 a transplant experiment was performed to assess the effect of freshwater migration distance on survival of coho salmon smolts. Fish were marked at the upstream sites and were separated into two groups, experimental and control. Experimental fish were transported downstream from their capture sites to a site three kilometers upstream from the main fence. Control fish were released at their capture sites. Eight separate transplants were performed; three from O'Connor Lake, two from Cub Creek, two from Misty Inlet and one from Long Lake.^Mean survival rates were calculated for the control and experimental fish and were compared using a two sample t-test.  76  Question 4: do fish survive differently with respect to habitat. Survival data were categorically coded by habitat and a regressed against survival. Survival estimates of laketype fish were coded as the value -1; stream-type habitats were coded as +1. A least-squares regression was performed on these data.  Question 5: does handling, anesthetizing and marking influence survival?  These questions were addressed in two separate experiments.^First, to assess the effects of handling the fish, one group was processed in the usual manner: trapped, anesthetized, measured, weighed, and marked. These fish were added to enclosures with fish which had only been trapped and not processed. Twelve enclosures were used with five handled and five control fish in each enclosure. Survival was monitored over a 7 day period then the experiment was terminated due to flood conditions. Survival data were compared using a paired t-test; second, a sample of 131 fish were processed in the usual manner and released at their upstream site. Survival data from these fish were compared to the survival of a sample of 130 fish which had been minimally handled and marked without anesthetic. Analysis was with ax 2 contingency table.  77  RESULTS:  Questions land 2: length at migration and timing of migration.  In all four systems, the stepwise regression procedure removed power terms from the multiple regression. These terms did not contribute to the regression any more than the length term alone. Response surfaces of the length time vs survival multiple regression (figures 29 - 32) show a general, negative effect of migration time, but no consistent effect by length at migration. Parameters included by the stepwise procedure differed between systems. For Misty Inlet, only the constant was left in the regression. In addition to the power terms, the stepwise procedure removed date and length variables; there were no statistical relationships between survival and timing or size (F-ratio = 0.35, p = 0.715, n = 11). The response surface in figure 29 showed little relationship between either of the independent variables and survival. Long Lake's date parameter was included with the constant but the regression was not significant (F-ratio = 4.916, p = 0.091, n = 6). The response surface (figure 30) demonstrated a trend with survival decreasing as the date of capture increased. There was less of a trend in the length - survival relationship. This regression's lack of  78  Figures 29 — 32.^Multiple regression response surfaces for the four site: Misty Inlet Creek (figure 29); Long Lake (figure 30); O'Connor Lake (figure 31); and Cub Creek (figure 32) in 1991. The y axis is percent .survival, the x axis is length in millimeters. The z axis is the middle day of each marking period. The x and z axes are displayed in a decreasing scale from the origin to aid in viewing the response surface.  79  Misty Inlet 1991  80 70 60  SURVIVAL^50 40 30  110  LZNGTR  80  Long Lake^1991  SURVIVAL  81  83  significance could be due to the small number of data points. The two independent variables were correlated (r = 0.699), which makes it difficult to draw conclusions from the response plane; it cannot be said which independent variable influenced survival. O'Connor Lake's date effect was significant (F-ratio = 12.788, p < 0.05, n = 14). The graph of the multiple regression (figure 31) displayed a very strong relationship between date and survival. The trend in survival vs capture date was similar to Long Lake; survival decreased as the season progressed. With respect to length, there was a trend suggesting large fish had lower survival^This may be due to the two independent variables being correlated (r = 0.846) and the size of captured fish increasing throughout the season. The Cub Creek multiple regression included both length and date parameters and was significant (F-ratio = 20.687, p = 0.001, n = 12) and the two independent variables were not correlated (r = -0.352). In this case there appears to be decreased survival with increasing capture date and decreased survival with increasing size (figure 32).  Question 3: distance of migration.  There was a significant difference in survival between experimental (migration distance = 3 km) and control fish  84  (migration distance > 10 km). Experimental fish displayed higher survival rates (0.51, 0.0, 0.64, 0.09, 0.50, 0.45, 0.40, 0.51) than the control fish (0.52, 0.0, 0.57, 0.08, 0.0, 0.44, 0.18, 0.26) (paired t-test, p < 0.05, df = 7).  Question 4: juvenile habitat.  The regression of habitat vs survival (figure 33) was significant (F-ratio = 5.05, p < 0.05). Fish from stream systems had higher survivals. The twenty three survival rates from the stream systems (Cub and Misty) and the twenty survival rates from the lake systems (O'Connor and Long) are listed in table 2.1.  Question 5: effects of handling.  The enclosure experiment displayed little difference between marked and control fish. All 10 fish in one enclosure died from asphyxiation when the enclosure drifted to low water during flood conditions. The remaining 11 enclosures had only one mortality, a marked fish, before the experiment was terminated.^There was no statistical difference between treatments (paired t-test, p = 0.338, df = 11) The results of the second experiment (testing the effects of anesthetic and handling) were not statistically  85  Figure 33. Linear least squares regression of habitat vs survival. Percent survival is on the y axis; habitat coded as —1 for lake fish and +1 for stream fish is on the x axis.  87  significant (X 2 = 1.408, df = 4, p = 0.843). Of the 131 non-anesthetized fish, 12 survived. Of the 130 anesthetized fish, 18 survived.  DISCUSSION:  The stepwise regression procedure indicated a negative effect of timing of migration in three of the four systems. Of these three, the two statistically significant effects were for the upper-watershed systems. For Long Lake (the third system), the regression was not significant at a=0.05. Significance in the two upper watershed systems may be due to their location or simply to larger sample sizes. The transplant experiment indicated that migration distance had an effect on survival. Since mortality rates were higher for upstream systems, this could result in the factors influencing mortality being more apparent. The multiple regression for Cub Creek indicated size was also a factor in survival. The size term was included in the model and displayed a negative coefficient indicating survival was inversely related to size. This is the only regression where the size term was significant, although O'Connor Lake displayed a similar but not statistically significant trend (p = 0.087). Once again, it was the upper systems displaying consistent trends which could be due to either larger sample sizes or greater  88  mortality pressures. Misty Inlet did not have significant regression coefficients for either the date or length effects. An examination of its response surface indicated there was a trend towards lower survival as the date of capture increased. This trend was evident in all four systems. It appears survival was highest early in the year and decreased as the season progressed. This is in agreement with Cameron (1958) who observed decreasing survival with increasing date of capture in his study of migrating pink salmon. The effect was clear in the systems where there was no correlation between length and timing; the result was less convincing where length and time were correlated. The effect of length on survival was difficult to ascertain, and no consistent patterns were observed. Bilton et al. (1982) manipulated weight and timing of release of hatchery reared smolts to examine smolt to adult survival. They found survival to adult stage was highest for smolts weighing 25.1 grams released on day 173. This study indicates survival during the smolt stage was highest at the start of the year and decreased throughout the year. High survival during the smolt migration may not result in high smolt to adult survival Results of the transplant experiment revealed survival was related to migration distance; fish with shorter migration distances survived better. This finding is also in agreement with Cameron (1958) who found increasing  89  survival with shorter migration distances. Effects of distance are of interest since there is no apparent relationship between size and survival. Larger fish have the ability to swim faster; chapter 2 established that the larger O'Connor Lake fish swam faster than the smaller Cub Creek fish.^All other factors being equal, large fish should have a shorter migration time and a higher survival rate if survival is related to time at risk to predation factors rather than migration distance alone. It appears some other factor, perhaps higher rates of predation on larger fish, swamps the effect of swimming speed. Alternatively, survival may be determined by the number of mortality agents encountered during migration rather than the duration of exposure to each agent. This could occur if the mortality agents are not overly mobile and kill a percentage of the prey which pass them, independent of swimming speed. Assuming my literature search was adequate, the effect of habitat on survival had not been examined prior to this thesis. Results of the regression indicated fish from river systems survived better than fish from lake systems. It has been documented that river-type and lake-type coho have different morphologies and behaviour (Swain and Holtby 1989); these differences appear to be correlated with survival. There may be advantages to living in an environment prior to migrating in it. River-type fish may be better at foraging or predator avoidance during the  90  smolt migration. The observed low survival rates (table 2.1) appear too extreme; initial reaction has been to credit high mortality to the effects of handling. Two experiments to examine the effects of handling on survival did not indicate mortality was increased due to handling, but these tests are open to criticism. For example, the effects of the enclosures are not fully known and the experiment was terminated after only seven days due to flood conditions. However, a number of authors have documented low survival rates in different species of salmonids during downstream migration. Cameron (1958) calculated the mortality of migrating pink salmon to be between 53% and 82%. Rogers et al. (1972) calculated the mortality of sockeye smolts (0. nerka) due to a single factor, predation by arctic char (Salvelinus alpinus), to be 32% and 51% in two separate systems. Wood (1987 b) found a relationship between smolt density and maximum daily mortality due to mergansers (figure 34). Considering the smolt densities observed in the Keogh river, the observed mortality rates are in line with this relationship. This statement would be more convincing had I a measure of merganser density. Holtby and Hartman (1982) is perhaps the most extensive examination of population dynamics of wild Coho salmon to date. Over a ten year period, they found very low survival rates from the period between September 1 and  91  Figure 34. (From Wood 1987) Relationship between percentage mortality by mergansers and smolt density in the Big Qualicum hatchery system in 1981. Dots indicate daily estimates of maximum mortality; solid line indicate expected relationship if merganser density remained constant.  93  the smolt enumeration, averaging 21% (SE =  0.016)  over a  ten year period. They attributed the low survival rates to high winter_mortality.^This study indicates the effects of winter mortality may be overestimated; a proportion of the mortality previously attributed to winter mortality may actually occur during the smolt migration.  CONCLUSION:  Of the factors examined, timing of migration and distance of migration, and juvenile habitat appeared to influence the survival of downstream migrating coho salmon smolts. Survival decreased as the date of capture increased and decreased as distance of migration increased. Lower survival was observed for lake - type fish. There was no relationship between length at migration and survival and there was no evidence of decreased survival due to handling.  94  CHAPTER 4: GENERAL CONCLUSIONS  The focus of this thesis was to examine mortality of downstream migrating coho salmon smolts. In comparison to the other coho life history stages, mortality during this stage had not been extensively examined. Mortality rates observed during the migration were higher than the anticipated. This result may alter the traditional views of coho freshwater survival. The current theory is severe freshwater mortality occurring in winter and relatively minor mortality occurring during the smolt migration (Hoitby and Hartman 1982). This research disputes that theory. Factors were found to influence survival: juvenile rearing habitat, timing of migration, and distance of migration, however, I can only speculate about the mechanisms which cause differential survival. An opportunity exists to study the mortality mechanisms of coho salmon smolts. No relationship was found between median length at migration and survival but chapter 2 established size selective mortality did occur. A single size was not selected for nor was a trend favoring smaller or larger fish found. It appears the center of the distribution had higher survival than the tails, independent of absolute length. Once again, I can only speculate to the mechanisms which cause this result. This thesis provides a basis for  95  further research on this little studied aspect of coho life history.  96 REFERENCES  Beamesderfer, R.C., B.E. Rieman, L.J. Bledsoe, and S. Vigg. 1990. Management implications of a model of predation by a resident fish on juvenile salmonids migrating through a Columbia River reservoir. North American Journal of Fisheries Management 10:290-304. Bilton, H.T., D.F. Alderdice, and J.T. Schnute. 1982. Influence of time and size at release of juvenile coho salmon (Oncorhynchus kisutch) on returns at maturity. Can. J. Fish. Aquat. Sci. 39: 426-447. Bley, P.W., 1987. Age, growth, and mortality of juvenile Atlantic salmon (Salmc salar) in streams: a review. U.S. Fish Wild. Surv., Biol. Rep. 87(4). 25pp. Cameron, W.M. 1958. Mortality during the Freshwater existence of the Pink Salmon. Fisheries Research Board of Canada Manuscript Report Series (Biological) No.669. Chambers, J. M., W.S. Cleveland, B. Kleiner, and P.A. Tukey. 1983. Graphical methods for data analysis. Wadsworth International Group, Belmont CA and Duxbury Press, Boston, MA. 395 p. Conlin, K. and B.D. Tutty. 1979. Juvenile Salmon Field Trapping Manual. Fisheries and Marine Service Manuscript Report #1530. Department of Fisheries and Oceans, Vancouver, B.C. Elson, P.F. 1962. Predator-prey relationships between fish-eating birds and Atlantic salmon. Fisheries Research Board of Canada Bulletin 133. Folmar, L.C., and W.W. Dickhoff. 1980. The parr-smolt transformation (smoltification) and seawater adaptation in salmonids. A review of selected literature. Aquaculture 21:1-37. Hart, P,J.B. and T.J. Pitcher. 1969. Field trials of fish marking using a jet inoculator. J. Fish Biol. 1:383385. Hartman, G.F. [ed]. 1982. Proceedings of the Carnation Creek Workshop, a 10 year review. Pacific Biological Station. Nanaimo, B.C.  97  Hartman, G.F. and J.C. Scrivener. 1986. Some strategy considerations for small stream restoration and enhancement with special emphasis on high rainfall streams such as Carnation Creek. p. 69-84. in J.H. Patterson [ed] Proceedings of the workshop on habitat improvements, Whistler, B.C. Can. Tech. Rep. Fish. Aquatic Sci. 1483, 219 p. Healey, M.C. 1982. Timing and relative intensity of sizeselective mortality of juvenile chum salmon (Oncorhynchus keta) during early sea-life. Can. J. Fish. Aquat. Sci. 39:952-957. Herbinger, C.M., G.J. Newkirk, and S.T. Lanes. 1990. Individual marking of Atlantic salmon: evaluation of cold branding and jet injection of Alcian Blue in several fin locations. J. Fish Biol. 36:99-101. Heggenes, J. and R. Borgstrom. 1988. Effect of mink, Mustela vison Schreber, predation on cohorts of juvenile Atlantic salmon, Salmo salar L., and brown trout, S. trutta L., in three small streams. J. Fish. Biol. (1988) 33,885-894. Hoar, W.S. 1953. Control and Timing of Fish Migration. Biol. Rev. 28:437-452. Holtby, L.B. and G.F. Hartman. 1982. The population dynamics of coho salmon (Oncorhynchus kisutch) in a west coast rain forest stream subjected to logging. In, G. Hartman [ed] Proceedings of the Carnation Creek workshop , a 10 year review. Pacific Biological Station. Nanaimo, B.C. Holtby, L.B., T.E. McMahon, and J.C. Scrivener. 1989. Stream temperatures and inter-annual variability in the emigration timing of coho salmon (Oncorhynchus kisutch) smolts and fry and chum salmon (0. keta) fry from Carnation Creek, British Columbia. Can. J. Fish. Aquat. Sci. 46: 1396-1405. Hunter, J.G. 1959. Survival and production of pink and chum salmon in a coastal stream. J. Fish. Res. Board Can. 16:835-886. Irvine, J.R. and B.R. Ward. 1989. Patterns of timing and size of wild coho salmon (Oncorhynchus kisutch) smolts migrating form the Keogh River Watershed on northern Vancouver Island. Can. J. Fish. Aquat. Sci. 46: 1086-1094.  98  Johnston, N.T., C.J. Perrin, P.A. Slaney, and B.R. Ward. 1990. Increased juvenile salmonid growth by stream fertilization. Can. J. Fish. Aquat. Sci. 47: 862-879. Johnston, N.T., J.R. Irvine, and C.J. Perin. 1987. Coho salmon- (Oncorhynchus kisutch) utilization of tributary lakes and streams in the Keogh River drainage, British Columbia. Can. MS Rep. Fish. Aquat. Sci. 1973: 54 p. Kelley, W.H. 1967. Marking of freshwater and marine fish by injected dyes. Trans. Amer. Fish. Soc. 96(2):163175. Larsson, P-0. 1985. Predation on migrating smolt as a regulation factor in Baltic salmon, Salmo Salar L., populations. J. Fish. Biol. 26,391-397. McMahnon, T.E. and L.B.Holtby. 1992. Behaviour, Habitat use, and movements of coho salmon (Oncorhynchus kisutch) smolts during seaward migration. Can. J. Fish. Aquat. Sci. 49:1478-1485. Mace, P.M. 1983. Predator-prey functional responses and predation by staghorn sculpins (Leptccottus armatus) on chum salmon fry (Oncorhynchus keta). PhD. thesis. University of British Columbia, Vancouver, B.C. Murphy, M.L., J.F. Thedinga, and K V. Koski. 1988. Size and diet of juvenile Pacific salmon during seaward migration through a small estuary in southeast Alaska. Fish. Bull. 86:213-222. Neave, F. 1953. Principles Affecting the Size of Pink and Chum Salmon Populations in British Columbia. J. Fish. Res. Bd. Can., 9(9):450-491 Parker, R.R. 1968. Marine mortality schedules of pink salmon of the Bella Coola River, central British Columbia. J. Fish. Res. Board Can. 25:757-794. Parker, R.R. 1962. Size selective predation among juvenile salmonid fishes in a British Columbia inlet. J. Fish. Res. Bd. Canada 28: 1503-1510. Post, J.R., and D.O. Evans. 1989. Size-dependent overwinter mortality of young-of-the-year yellow perch (Perca flavescens): laboratory, in situ enclosure, and field experiments. Can. J. Fish. Aquat. Sci. 46:19581968. Rieman, B.E. and R.C. Beamesderfer. 1991. Estimated loss of juvenile salmonids to predation by northern squawfish,  99 walleyes and smallmouth bass in John Day Reservoir, Columbia River. Tran. Amer. Fish. Soc. 120:448-458. Rogers, D.E., L. Gilbertson, and D. Eggers. 1972. Predator-prey relationship between arctic char and sockeye salmon smolts at the Agulowak River, Lake Aleknagik, in 1971. University of Washington, College of Fisheries, Fisheries Research Institute, Circular No. 72-7. Ruggerone, G.T., and D.E. Rogers. 1984. Arctic char predation on sockeye salmon smolts at Little Togiak River, Alaska. U.S. National Marine Fisheries Service Fishery Bulletin 82:401-410. Ruggles, C.P. 1980. A review of the downstream migration of Atlantic salmon. Can. Tech. Rep. Fish. Aquat. Sci. No. 952. Shapovalov, L. and A.C. Taft. 1954. The life histories of the steelhead rainbow trout (Salmo gairdneri gairdneri) and silver salmon (Oncorhynchus kisutch). California Dept. of Fish and Game Fish Bull. No. 98. Swain, D.P., and L.B. Holtby. 1989. Differences in morphology and behavior between juvenile coho salmon (Oncorhynchus kisutch) rearing in a lake and in its tributary stream. Can. J. Fish. Aquat. Sci. 46: 14061414. Swales, S., F. Caron, J.R. Irvine, and C.D. Levings 1988. Overwintering habitats of coho salmon (Oncorhynchus kisutch) and other juvenile salmonids in the Keogh River system, British Columbia. Can. J. Zool. 66: 254261. Thedinga, J.F. and K V. Koski. 1984. A stream ecosystem in an old-growth forest in southeast Alaska. Part VI: the production of coho salmon, Oncorhynchus kisutch, smolts and adults form Porcupine Creek. In Meehan, W.R., T.R. Merrell, Jr., and T.A. Hanley [ed] Fish and Wildlife Relationships in Old-Growth Forests. American Institute of Fishery Research Biologists. 1984. Tukey, J.W. 1977. Exploratory Data Analysis. AddisonWesley Publishing Company, Redding, Massachusetts. 688 p. Wood, C.C. 1987. Predation of juvenile Pacific salmon by the common merganser (Mergus merganser) on eastern Vancouver Island. I: Predation during the seaward migration. Can. J. Fish. Aquat. Sci. 44:941-949.  100  Wood, C.C. 1987 b. Predation of juvenile Pacific salmon by the common merganser (Mergus merganser) on eastern Vancouver Island. II: Predation of stream-resident juvenile salmon by merganser broods.^Can. J. Fish. Aquat.,Sci. 44:950-959.  


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items