Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The northern fur seal : biological relationships, ecological patterns and population management Trites, Andrew W. 1990

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

Item Metadata

Download

Media
831-UBC_1991_A1 T74.pdf [ 17.26MB ]
Metadata
JSON: 831-1.0302419.json
JSON-LD: 831-1.0302419-ld.json
RDF/XML (Pretty): 831-1.0302419-rdf.xml
RDF/JSON: 831-1.0302419-rdf.json
Turtle: 831-1.0302419-turtle.txt
N-Triples: 831-1.0302419-rdf-ntriples.txt
Original Record: 831-1.0302419-source.json
Full Text
831-1.0302419-fulltext.txt
Citation
831-1.0302419.ris

Full Text

The Northern Fur Seal: Biological Relationships, Ecological Patterns and Population Management By Andrew W. Trites B. Sc. McGill University M. Sc. University of British Columbia A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE STUDIES ZOOLOGY We accept this thesis as conforming to the required standard  THE UNIVERSITY OF BRITISH COLUMBIA December 1990 © Andrew W. Trites, 1990  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.  Department The University of British Columbia Vancouver, Canada Date  DE-6 (2/88)  Abstract Data collected from Pribilof far seals,  Callorhinus ursinus, on land (1911-89) and  at sea (1958-74) are analyzed to establish biological relationships and distinguish ecological patterns that are relevant to understanding and managing northern fur seal populations. The thesis follows the development of the fur seal from conception and birth through to sexual maturity and finally to a synthesis of the earlier material in terms of population regulation, management, and reasons for the decline of the Pribilof herd. Growth curves show that male fetuses grow faster and larger than female fetuses, and that fetal size is influenced by the age, size, and reproductive history of the mother. Juvenile and adult fur seals experience pronounced seasonal increases and decreases in body length and mass. Rapid gains in mass and growth occur during a brief 1-3 month period as the population migrates through the coastal waters of northern British Columbia and Alaska on its way to the Pribilof Islands. Body mass is gradually lost during the rest of the year while fasting on land and wintering along the coasts of Washington, Oregon, and California. The timing of migration and pupping is highly synchronized from year to year and may be related to the effect of climatic conditions on pup survival during the breeding season. Predictions from a thermal budget developed for pups and the results of a seasonal decomposition of weather patterns on the Pribilof Islands show that the synchronism of births in early July corresponds to the start of three months of conditions that are optimal for growth and survival of pups. Long term fluctuations are noted in  ii  pup mass and subadult growth rates which may be related to underlying, large scale natural changes in prey abundance. Changes in the physiological condition (body growth) and vital rates (survival and reproduction) are analyzed for the period 1911-89 as the population increased and decreased. Few density dependent relationships could be demonstrated. Two hypotheses concerning the current decline of the Pribilof population are reviewed and a new, third hypothesis is proposed. The thesis also examines biases in data collection related to the effects of tagging and the handling of fur seals and outlines some directions for future research.  iii  Contents Abstract  ii  List of Tables  xiv  List of Figures  xxii  Acknowledgement 1  2  xxiii  General Introduction  1  1.1  Historical Background  3  1.2  Thesis Overview  6  1.3  Survey of Individual Chapters  7  Fetal Growth: Life History Strategy and Sources of Variation  11  2.1  Introduction  11  2.2  Biology of Fur Seal Reproduction  12  2.3  Materials and Methods  13  2.3.1  Verification and Detection of Biases in Data Collection . . . .  14  2.3.2  Average Fetal Growth  14  2.3.3  Differences Between Primiparous and Multipaxous Females  2.3.4  Effect of Maternal Age on Fetal Size  2.4  Results and Discussion  .  15 16 16  iv  2.5  2.6 3  2.4.1  Verification of Data  16  2.4.2  Age and Parity Biases  19  2.4.3  Sex Ratio Biases  23  2.4.4  Average Fetal Growth  24  2.4.5  Differences Between Primiparous and Multipaxous Females  2.4.6  Effect of Maternal Age on Fetal Size  .  28 29  Overview  30  2.5.1  Observed and Predicted Neonatal Size  31  2.5.2  Length of Gestation  32  2.5.3  Maternal Resources  33  2.5.4  Conclusion  33  Summary  34  Changes in Fetal Growth and the Condition of Pregnant Females from 1958-72  35  3.1  Introduction  35  3.2  Materials and Methods  36  3.2.1  Female Condition Index  36  3.2.2  Annual Fetal Growth  37  3.3  3.4  3.5  Results  38  3.3.1  Female Condition Index  38  3.3.2  Annual Fetal Growth  39  Discussion  40  3.4.1  Female Condition  40  3.4.2  Food Availability and Fetal Growth  43  3.4.3  Changes in the Fur Seal Food Base  45  Conclusion  47  v  3.6 4  5  Summary  47  Reproductive Synchrony and the Estimation of Mean Date of Birth from Daily Counts of Pups  49  4.1  Introduction  49  4.2  Methods  50  4.2.1  Sigmoid Method  50  4.2.2  Summation Method  51  4.2.3  Pup Mortality  53  4.2.4  Pur Seal Pup Counts  54  4.3  Results  55  4.4  Discussion  61  4.5  Summary  63  Thermal Budgets and Climate Spaces: The Impact of Weather on Pup Survival  64  5.1  Introduction  64  5.2  Biology of the Pup  66  5.3  Methods  68  5.3.1  Thermal Budget  68  5.3.2  Available Radiant Energy  69  5.4  Parameterization  70  5.4.1  Mass and Diameter of the Pup  70  5.4.2  Body Core Temperatures  71  5.4.3  Metabolic Rates  72  5.4.4  Heat Transfer Resistance  73  5.4.5  Surface Emissivity and Evaporative Heat Loss  74  5.4.6  Environmental Conditions  74  vi  75  5.4.8  Radiant Energy Available  76  Results  79  5.6  Discussion  80  5.6.1  Weather and Pup Mortality  83  5.6.2  Behavioural Modifications  85  5.6.3  Role of Body Size  86  5.6.4  Future Research  .  Summary  87 88  5.A Appendix: Sensitivity Analysis  89  The Influence of Weather on the Fur Seal's Life Cycle  92  6.1  Introduction  92  6.2  Methods  93  6.3  Results and Discussion  94  6.3.1  Mean Monthly Conditions  94  6.3.2  Seasonal Decomposition  94  6.3.3  Impact of Weather on Seals  97  6.4 7  Thermal Budgets  5.5  5.7  6  5.4.7  Summary  102  Does Tagging and Handling Affect the Growth of Pups?  103  7.1  Introduction  103  7.2  Materials and Methods  104  7.2.1  Tagging, Marking & Weighing  104  7.2.2  Analysis of Growth Patterns  105  7.3  Results and Discussion  106  7.3.1  Growth Rates  108  7.3.2  Other Species  113 vii  7.4 8  115  Biased Estimates of Pup Mass: Origins and Implications  116  8.1  Introduction  116  8.2  Milk Consumption  117  8.3  Seasonal Pup Growth  120  8.3.1  Sample Size Effects  124  8.3.2  Overview  128  8.4 9  Summary  Summary  128  Food Abundance and Annual Fluctuations in the Growth of Pups  130  9.1  Introduction  130  9.2  Materials and Methods  132  9.2.1  Weighing  132  9.2.2  Annual Variations in Weight  133  9.3  Results and Discussion  134  9.4  Summary  141  10 Seasonal Growth Fluctuations During the Fur Seal Migration  142  10.1 Introduction  142  10.2 Methods  143  10.2.1 Fur Seal Biology  143  10.2.2 Collection of Data  144  10.2.3 Analysis of Data  146  10.3 Results  148  10.3.1 Seasonal Fluctuations  148  10.3.2 Allometric Relationships  154  viii  10.3.3 Growth and Migration  155  10.4 Discussion  158  10.4.1 Pinniped Growth  158  10.4.2 Northern Fur Seals  161  10.4.3 Indeterminate Growth  163  10.4.4 Fluctuations in Body Length  164  10.4.5 Growth and Feeding Location  167  10.4.6 Future Considerations  168  10.5 Summary  170  11 Interannual Variability in the Growth of Adults and Juveniles from 1958-74  171  11.1 Introduction  171  11.2 Methods and Results  172  11.2.1 growth curves  174  11.2.2 size of mature non-pregnant females  177  11.2.3 growth rates of immature females  178  11.2.4 condition indices  182  11.3 Discussion  184  11.4 Summary  190  12 Haulout Composition, Capture Efficiency and Escapement During the Harvest  191  12.1 Introduction  191  12.2 Background to the Harvest  192  12.3 Study Area and Methods  194  12.4 Results and Discussion  196  12.4.1 Composition of the Haulouts ix  196  12.4.2 Capture at the Haulouts  200  12.4.3 Escapement  203  12.4.4 Future Management Considerations  .  12.5 Summary  205 205  13 Changes in Abundance and Body Size in Subadult Males  207  13.1 Introduction  207  13.2 Background to the Harvest  208  13.3 Origins & Implications of Sampling Biases  210  13.3.1 Measuring Techniques  210  13.3.2 Size Limit Effects  212  13.3.3 Season Length Effects  214  13.3.4 Tagging Effects  217  13.4 Mean Length of Total Harvest, 1933-58  220  13.5 Mean Length of 3 & 4 y olds, 1915-83  223  13.6 Discussion  226  13.7 Summary  228  14 Changes in Reproductive Rates and the Mean Age at First Reproduction During the Herd Decline, 1957-74  230  14.1 Introduction  230  14.2 Pregnancy Rates  232  14.3 Proportion of Muciparous Females  234  14.4 Age at First Reproduction  236  14.5 Discussion  238  14.6 Summary  243  x  15 Density Dependent Responses and the Management of Pinniped Populations: A Case Study  244  15.1 Introduction  244  15.2 The Fur Seal Data Base  247  15.3 Growth and Body Size  249  15.4 Reproduction  257  15.5 Survival  259  15.6 Discussion  261  15.7 Summary  265  16 Northern Fur Seals: W h y are They Declining?  266  16.1 Introduction  266  16.2 Entanglement Related Mortality  267  16.3 Commercial Fisheries  271  16.4 Fur Seal Diet  274  16.5 A Depleted Food Base?  277  16.6 The Bottle Neck Hypothesis  278  16.7 Decline of Other Species  281  16.8 Future Research  285  16.9 Summary  288  17 General Discussion  290  17.1 Biological Relationships and Ecological Patterns  291  17.2 Population Management  293  Literature Cited  297  xi  List of Tables 1.1  Population size of northern fur seals on the breeding islands  2.1  Number of fetuses collected at sea by month from 1958-74  19  2.2  Three growth models representing the length of fetuses  27  3.1  Changes in the mean mass of mothers and pups weighed at term . .  38  3.2  Linear regression coefficients for fetal growth in length  43  3.3  Linear regression coefficients for fetal growth in mass  44  4.1  'Sigmoid' estimates of the median birth date for pups born on 4 study sites  4.2  56  'Summation' estimates of the mean birth date for pups born on 4 study sites  5.1  58  Average daily weather conditions recorded at Saint Paul during July and at the Galapagos during September  5.2  5.4  75  Parameters used to estimate the upper and lower physiological limits of northern fur seal pups and Galapagos fur seal pups  5.3  3  76  Estimates of resistance used to determine the upper and lower physiological limits of fur seal pups  77  List of symbols and values  78  xii  5.5  Parameter estimates and solutions of the partial differential equation used to determine the total error in the intercept of the energy budget equation  6.1  90  Average surface temperature of the Bering Sea by month from 1953 to 1982  7.1  101  Growth rates estimated for tagged and untagged pups weighed in 1962 and 1957-66  8.1  Ill  Estimated density of northern fur seal milk determined from the specific density and percent composition of fur seal milk components . . 118  8.2  Mean mass of male and female pups weighed between Aug 29 - Sep 3 122  10.1 Mean length and mass of males, non-pregnant females and pregnant females by age  158  11.1 Number of males collected at sea by month from 1958-74  172  11.2 Number of pregnant and post-partum females collected at sea by month from 1958-74  173  11.3 Number of non-pregnant females collected at sea by month from 1958-74  173  13.1 Mean length, standard deviation (sd) and sample size of subadult males harvested until July 27 of each year from 1933-58 on the Pribilof Islands  221  13.2 Mean length, standard deviation, and sample size of 3 and 4 y olds harvested on St. Paul Island  224  xiii  14.1 Number of female northern fur seals captured in 1959 and 1960 that were nulliparous, primiparous and multiparous, and the proportion that were currently or previously pregnant  238  15.1 Pinnipeds observed to exhibit density dependent changes body growth, age at sexual maturity, pregnancy rates, juvenile survival and adult survival  245  16.1 Abundance and seal diet composition estimates for major fish species consumed by northern fur seals in the Bering Sea and Gulf of Alaska 275 16.2 Estimated annual consumption of walleye pollock in the eastern Bering Sea  276  xiv  List of Figures 1.1  Breeding islands of the northern fur seal  2  2.1  Length and mass of 7003 fetuses by date of sampling  18  2.2  Natural logarithm of fetal mass by date of sampling  20  2.3  Number of primiparous and multiparous females captured by age from 1958-74  21  2.4  Biases in monthly samples  22  2.5  Maternal size and the effect of maternal age on fetal size in multiparous females  25  2.6  Growth curves for male and female fetuses  • • • 26  2.7  Mean length of multiparous females collected from December to June  30  3.1  Female condition versus the date when collected  39  3.2  Mean condition of multiparous females carrying male and female fetuses 40  3.3  Linear regression of fetal length and mass over time  3.4  Growth curves for male and female fetuses taken from multiparous  41  females aged 8-13 y  42  3.5  Estimated size of the fetus at term from 1958-72  46  4.1  Daily number of live pups counted on 4 study sites from June 15 to August 5  57  xv  4.2  Relative number of pups born each day on 4 study sites from June 15 to August 5  4.3  59  Cumulative total of dead pups counted on a portion of Vbstochni rookery in 1951  5.1  60  Climate space diagrams for northern fur seal pups on the Pribilof Islands, Alaska  5.2  80  Climate space diagrams for northern fur seal pups on San Miguel Island, California  5.3  81  Climate space diagrams of a northern fur seal pup on the Pribilof Islands contrasted with that of a Galapagos fur seal pup on the Galapagos Islands  6.1  82  Smoothed counts of territorial bulls, nursing females, and pups made by Peterson (1965) at Kitovi study site in 1962 and 1963  6.2  93  Mean weather conditions (air temperature, wind speed, and relative humidity) recorded by month at St. Paul Island for the period 1956 to 1986  6.3  95  Seasonal decomposition of monthly air temperatures recorded on St. Paul Island from 1956-68  6.4  96  Seasonal decomposition of monthly wind speeds recorded on St. Paul Island  6.5  97  Seasonal decomposition of monthly levels of relative humidity recorded on St. Paul Island from 1956-86  6.6  Seasonal component of air temperature, wind speeds and relative humidity levels from 1956-86  6.7  98  99  The irregular component of air temperatures, wind speeds and relative humidity levels from 1956-86  xvi  100  7.1  Annual differences in the mean body mass of tagged and untagged, male and female pups weighed on four rookeries: Zapadni Reef, Northeast Point, Polovina and Reef  107  7.2  Mean mass of tagged pups versus those without tags  108  7.3  Mean mass of the tagged, marked, and untagged/unmarked pups weighed on four rookeries on Sep 2-3, 1962  7.4  109  Mean body mass of males and female pups measured on four rookeries from September to October 1962  7.5  110  Mean body mass of tagged and untagged pups weighed from Aug 29 to Sep 3  7.6  112  Differences between the mean body size of tagged and untagged pups subsequent to tagging  8.1  113  Growth and milk consumption for male and female pups during one day of feeding  119  8.2  Relationship between the mass of male and female pups  121  8.3  Mean body mass of untagged pups weighed from Aug 29 - Sep 3  8.4  Relationship between mean body size and the number of untagged  . . 123  pups weighed on three rookeries: Zapadni Reef, Polovina and Reef . 125 8.5  Relationship between mean body size of a sample of tagged pups and the total number of pups bearing tagged  9.1  127  Mean mass of untagged pups weighed over a two day period from 1957-71  135  9.2  Mean mass of tagged pups weighed over a two day period from 1957-66136  9.3  Estimated number of 3-year old walleye pollock in the eastern Bering Sea from 1964-88  138  xvii  9.4  Mean mass of tagged and untagged pups weighed on four rookeries from 1957-71  140  10.1 The seven regions where fur seals were collected from 1958-74 . . . . 145 10.2 Length and mass of 2 008 males by age in days 10.3 Length and mass of 6 493 non-pregnant females by age in days  149 . . . 150  10.4 Length and mass of 9630 pregnant and postpartum females by age in days  151  10.5 Seasonal changes in the growth of immature males and females . . . 152 10.6 Seasonal changes in the size of pregnant and non-pregnant females . 153 10.7 Smoothed growth curves generated with lowess using large / values that remove seasonal growth  fluctuations  155  10.8 Comparison of male, pregnant female and non-pregnant female growth curves  156  10.9 Date and location where fur seals of different ages and reproductive statuses were collected  157  10.10 Comparison of the timing of migration through the North Pacific for different components of the fur seal herd  159  10.11 Comparison of seasonal changes in the size of immature males, immature females, pregnant females and non-pregnant females  160  11.1 Growth curves for immature males and females (ages 1.5 - 4.5 y) collected from January through April over three time periods: 195862, 1963-68 and 1969-74  175  11.2 Growth curves for pregnant and non-pregnant females (ages 4.5+ y) collected from January through April over three time periods: 195862, 1963-68 and 1969-74  176  xviii  11.3 Mean length and mass of mature non-pregnant females aged 14.5 y +  collected from 1958-74  179  11.4 Growth rates of immature females  181  11.5 Mean conditions of immature females by year of collection  183  11.6 Mean condition of mature non-pregnant females aged 14.5 y col+  lected from 1958-74  184  11.7 Fluctuations in pollock recruitment and commercial catches of salmon, Pacific cod and Dungeness crabs in the North Pacific Ocean  187  11.8 Maximum, average, and minimum monthly bottom temperature anomalies in three areas in the Bering Sea from 1950 to 1982  188  11.9 Solar activity from 1955-85  189  12.1 Rookery and hauling ground sites on St. Paul Island  193  12.2 Number of bulls and bachelors present prior to each harvest at Little Zapadni and Lukanin during the month of July from 1980-83 . . . . 196 12.3 Standardized number of bulls and bachelors present prior to each harvest at Little Zapadni and Lukanin from June 30 to August 4 . . 197 12.4 Number of bulls versus bachelors using the haulouts on the day of harvest at Little Zapadni and Lukanin from 1980-83  199  12.5 Percentage of hauled bulls and bachelors captured by the sealers at Lukanin and Little Zapadni from 1980-83  201  12.6 Number of hauled bulls present versus the rate of capture of bachelors at Little Zapadni and Lukanin from 1980-83  203  13.1 Body length limits used to control the harvest of male fur seals . . . 210 13.2 Relationship between body length and tail length for 117 subadult males randomly sampled from the St. Paul commercial harvest in 1983212  xix  13.3 Comparison of the distribution of measured lengths in 1950 from tagged 3 y olds with the lengths of individuals that lost their tags  . 213  13.4 Comparison of the distribution of measured lengths in 1951 from tagged 4 y olds with the lengths of individuals that lost their tags  . 214  13.5 Size distribution of subadult males harvested on St. Paul Island in 1980 and 1983  215  13.6 The influence of season length on the estimated mean size of males harvested on St. Paul Island in 1956  216  13.7 Mean length of tagged and untagged 3 y olds killed and measured over 7 rounds in 1963 on St. Paul Island 13.8 Harvest days on St. Paul and St. George Islands from 1933-58  217 . . . 218  13.9 Mean length of tagged and untagged 3 y olds killed and measured in 1962, 1963 and 1964  219  13.10Mean length of males harvested until July 27 of each year from 193358 on St. Paul and St. George Islands  220  13.11Inter-annual trend in mean length of males harvested until July 27 of each year from 1933-58 on St. Paul and St. George Islands  222  13.12Mean length of 3 and 4 y olds harvested on St. Paul Island since 1915 225 14.1 Numbers of pups born and bulls counted on St. Paul Island from 1911-89 . . .  231  14.2 Age specific pregnancy rates (including postpartum animals) of 15 020 seals collected at sea in all sampling areas during 1958-74  233  14.3 Age specific reproductive condition of 10 824 females ages 0-10 y collected at sea in all sampling areas during 1958-74  234  14.4 Annual pregnancy rates of females between the ages of 8-13 y collected from California to British Columbia and in the Pribilof region 235  xx  14.5 Proportion of 6 and 7 y old females collected in all areas that were multiparous in the 1958-74 pelagic samples  236  14.6 Mean age at first reproduction in 1959 to 1974  239  14.7 Ratio of pups born to the number of bulls counted on St. Paul Island 241 15.1 Number of pups born and bulls counted on St. Paul Island from 1911-89249 15.2 Fetal size at term, 1958-72  251  15.3 Mean mass of tagged and untagged pups weighed from August 29 to September 3 on St. Paul Island when approximately two months old 252 15.4 Mean lengths of subadult males harvested on St. Paul Island  . . . . 253  15.5 Mean tooth weights of 3 y olds males harvested on St. Paul Island . 255 15.6 Growth curves for non-pregnant females (ages 5.5 + y) collected from January through April  257  15.7 Annual growth rates of immature females between the ages of 1.5 and 4.5 y  258  15.8 Mean age at first reproduction and pregnancy rates of females between the ages of 8-13 y  259  15.9 The survival rate of pups on land (0-4 m) and during the next 20 months while at sea  260  16.1 Juvenile survival rates of male northern fur seals  267  16.2 Major fishing regions in the Bering Sea and Gulf of Alaska  271  16.3 Commercial catch of walleye pollock and Pacific ocean perch in the Gulf of Alaska and Bering Sea  272  16.4 Estimated numbers of 3 year old walleye pollock in the Bering Sea and Gulf of Alaska  274  16.5 Annual rate of decline in numbers of Stellar sea Hons in the Gulf of Alaska and southern Bering Sea xxi  282  16.6 Decline in the numbers of harbour seals and Stellar sea lions in the Gulf of Alaska  283  16.7 Counts of kittiwakes and murres on census plots on the Pribilof Islands285  xxii  Acknowledgement The northern fur seal data base was collected and prepared by Canadian and American biologists, as members of the North Pacific Fur Seal Commission. I am grateful to the Department of Fisheries and Oceans, Canada and to the National Marine Fisheries Service, U.S.A. for allowing me access to their data. The data set represents hundreds of thousands of hours of collection and preparation, and has involved the efforts of numerous individuals. During my 'data mining' field trips, I was warmly welcomed into the homes of Jacky Booth in Nanaimo, and Sandy O'Neil and Tom Quinn in Seattle. Their fine food and stimulating conversations made my stays enjoyable and I thank them for it. Michael Bigg in Nanaimo, and Laurie Briggs and Sherry Pearson at the National Marine Mammal Lab in Seattle were invaluable in helping me to obtain data files and to find written records in the fur seal archives. My analysis of the data could not have been accomplished without the computing facilities and support of the staff at the Bio-Sciences Data Centre at UBC. Thank you Alistair Blachford, Susan Ertis, Shirley Ludwig, Charles Mathieson and Joerg Messer and for your patience and insight into my computing needs. Many people commented on my research and writings or provided stimulating discussion on the results of my findings. In particular I would like to thank Michael Bigg, Monique Bournot-Trites, Laurie Briggs, Ian Boyd, Dennis Chitty, Simon Courtenay, Doug DeMaster, Bill Doidge, Dean Fisher, Chuck Fowler, Roger Gentry, Harry Joe, Hiro Kajimura, Peter Larkin, Dave Lavigne, Richard Laws, Don Ludwig, Mike Perez, Don Robinson, Al Roppel, Joe Scordino, Dolph Schluter, Vic-  xxiii  tor Scheffer, Tony Sinclair, Carol Thommasen, Carl Walters, Peter Watts, Norman Wilimovsky and Anne York. I am grateful to my research committee (Michael Bigg, Buzz Holling, Peter Larkin, Don Ludwig and Carl Walters) for their interest in my work and their support throughout the course of my studies. They guided my research and helped me to clarify my thoughts and expression of ideas. I am especially appreciative of the time spent with Don Ludwig and Mike Bigg and for the encouragement they have continuously given me. I have benefited from Don's sense of perfection and his intuitive mathematical skills. He has influenced my approach to data analysis and deepened my appreciation for graphical exploratory methods. Mike's insight into northern fur seals and other pinnipeds was invaluable to me. His ability to recognize the essence of the fur seal story and the excitement he made me feel about marine mammal research have had a major influence on me and on what I was able to achieve. To both individuals I offer my sincere thanks. Last on my list of acknowledgements, but first in my mind is my family. I am grateful to my parents, Ron and Lillian Trites, for instilling in me a curiosity about the world around me that continues to bring me joy and satisfaction in my research. My children, Daniel and Vincent, watched patiently from the sidelines as I worked on the computer late into the night at the foot of my youngest son's bed. They were also quick to understand when I was late getting them to and from their appointments and sport activities. Thanks to them I was able to keep a proper perspective on my work. My wife Monique has always been a source of strength and support to me. Her inquisitive mind and deep sense of love have been with me throughout and have provided the much needed balance in my life.  Mike Bigg died as I was preparing to submit my dissertation. It leaves me with great sadness. I wish to dedicate this thesis to his memory.  xxiv  Chapter 1  General Introduction Two small fog bound islands in the southern Bering Sea, known as the Pribilofs, contain the breeding grounds of approximately 70% of all northern fur seals, Cal-  lorhinus ursinus. This species is the most abundant of nine related fur seal species that survived the devastating exploitation by sealers during the 18th and 19th centuries. Demand for the luxurious velvet underfur reduced the original range of the northern fur seal to four breeding locations: the Kurile Islands, Robben Island, the Commander Islands and the Pribilof Islands (Fig. 1.1 and Table 1.1). Since then two recently formed breeding populations have been discovered on San Miguel Island in 1968 and on Bogoslof Island in 1980 (Peterson and Cooper 1968; Peterson et al. 1968b; Lloyd et al. 1981; Loughlin and Miller 1989). In addition to these breeding islands, northern fur seals have been reported to haul out on other islands in the eastern North Pacific Ocean and Bering Sea (Fiscus 1983). The central focus of this dissertation is the northern fur seal population breeding on the Pribilof Islands. Being the largest fur seal herd in the world, this population has been intensively hunted. As a consequence it has also been intensively studied as well. The following pages and chapters summarize a large body of scientific study done by many researchers and presents new analysis of data sets that span almost a century in some cases. By analyzing the historical data set, I attempt to establish  1  Chapter 1  General Introduction  2  Figure 1.1: Breeding islands of the northern fur seal: the Pribilof Islands (1. St. Paul Island, 2. St. George Island), the Commander Islands (3. Bering Island, 4. Medny Island), the Kurile Islands (5. Kamennye Lovushki, 6. Srednev Rocks), 7. Robben Island (Tyuleni), 8. Bogoslof Island and 9. San Miguel Island.  key biological relationships and distinguish ecological patterns that are relevant to understanding and managing northern fur seal populations. The fur seal data base contains a wealth of information and is, as I will attempt to show, a window into understanding the biology and dynamics of other pinniped populations. Fresh insights into the future are often gained by probing the past. Thus I begin by briefly reviewing the history of exploitation and scientific study of northern fur seals on the Pribilof Islands. This is followed by an overview of the dissertation and a general introduction to each chapter.  Chapter 1  General Introduction  3  Table 1.1: Population size of northern fur seals on the breeding islands.  Population Pribilof Islands Commander Islands Robben Island Kurile Islands San Miguel Island Bogoslof Island Total  1.1  Numbers  Reference  900 000 200 - 220 000 70 - 80 000 45 - 50 000 4 000 400  NPFSC 1984a NPFSC 1984a NPFSC 1984a NPFSC 1984a NPFSC 1984a Loughlin and Miller 1989  1254 400  Historical Background  Nineteen years after discovering the Pribilof Islands in 1786, the dwindling fur seal herds held little profit for the stockholders of the Russian American Company. The company tried to increase dividends by implementing conservation practices such as enforcing periodic hunting closures in 1806-07 and 1835-36 (Martin 1946; Roppel and Davey 1965; Hulley 1970). They even considered suspending the kill every fifth year in 1821 to ensure population growth (Scheffer et al. 1984). The company's monopoly over the resources of the North Pacific region enabled them to consider such actions. Later in 1837, they experimented with an idea to exempt all females from the hunt. This policy, which became permanent in 1848 (Martin 1946), probably provided a large sustainable harvest and restored the herd to the size observed by Captain Pribilof when he discovered the breeding islands in 1786. The Russian American Company managed the northern fur seal herds to ensure a sustainable profit based on their understanding of fur seal population biology. They observed for example that the seals were polygamous and that immature males hauled out separately from the breeding population that used the rookeries. This led to the concept that continues to be held today that subadult males are excess  Chapter 1  General Introduction  4  to the needs of the population and can be harvested without impeding population growth. They also observed that each female on the rookeries gave birth to a single pup. This enabled a Russian named Veniaminov to estimate the numbers of seals that would return over a 22 year period 1835-56. He predicted the population would increase five-fold before stabilizing. At this point the population could sustain an annual harvest of 32 000 males provided that one-fifth of the bachelor population was spared annually (Veniaminov 1840; Elliot 1875; cited by Scheffer et al. 1984). The Russians also had some understanding of the homing ability of the seals and the distinctiveness of each rookery and associated haulouts after a Russian overseer on St. Paul Island cut off the ears of some of the seals captured and released the young males in the early 1860s and in 1870. In the year following each experiment, many of the marked seals were recaptured on the same hauling ground (Elliot 1884; cited by Scheffer et al. 1984). Understanding about fur seal biology was largely ignored when Americans began harvesting fur seals after the sale of Alaska in 1867. Sealing ships financed by three groups of Californian merchants arrived in the first summer off the Pribilofs, worked out an arrangement among themselves, and began a reckless slaughter taking 280 000 skins in just one year compared to the past annual Russian kill of 17 820 (Andrews 1931). Excessive land harvests, combined with inefficient hunting of seals at sea from boats registered in Japan and Canada quickly decimated the herd such that by 1910 the herd that once numbered close to 3 million animals consisted of only 125 000. Private enterprise sealing was removed from the Pribilof Islands in 1911 and the United States government transferred responsibility for the seal harvest to the Department of Fisheries. A moratorium was placed on the pelagic harvest of fur seals in return for which Canada and Japan would receive 15% of all furs secured on land. Biologists with the Fisheries Department began assessing the fur seal stock  Chapter 1  General Introduction  5  in 1911 by counting all live and dead pups, harem and idle bulls, and by recording the lengths of all males killed. Pups were branded in 1912 in an attempt to relate age to length of kill in subsequent years, and a 5 year moratorium was placed on the commercial harvest starting in 1914. Re-instituting fur seal management based on the principles of fur seal biology enabled the herds to recover. In 1940, the concern of biologists was to build the herd and size of the commercial harvest of subadult males. However, observations concerning the size of the harems, numbers of pups born, and survival rates during the 1950s made it apparent that the herd was no longer increasing and perhaps was not very productive. For example, maximum reproductive rates were likely surpassed as harem sizes increased from 42 cows per bull in 1935 to 94 in 1947 (Kenyon et al. 1954). Large fluctuations (5 - 6%) in the mortality of pups on land were also interpreted as a sign that the fur seal herd was no longer growing (NPFSC 1962). A third indication of this trend appeared in the kill data. Despite large numbers of pups born the annual kill was below expectation. Using a combination of tagging studies and the recently developed tooth aging technique of Scheffer (1950), Chapman (1961) discovered large fluctuations in juvenile pelagic survival rates. Because of the apparent lower pup survival at high population densities, the hypothesis of density dependent survival was promoted and subsequently incorporated into management policy. The hypothesis regarding density dependent pup survival led Chapman (1961) and Nagasaki (1961) to conclude that reducing the population would make the herd more productive. Chapman suggested lowering the female population from 1200 000 to 800 000 animals. This was to produce 480 000 pups and ensure a maximum sustainable yield of 60 000 males and 30 000 females. Thus between 1956 and 1963 a harem raiding program was conducted that yielded 270 054 female furs, most of which were destroyed because no developed market existed to handle them (Lander 1980a). The estimated annual production of pups during the herd reduction period  Chapter 1  General Introduction  6  was 120 000 short of the predicted level. Surprisingly the slaughter of females continued despite this shortcoming. In response Chapman (1964) revised and reduced his estimate of maximum sustainable yield to 581 000 reproducing females and later to 471 000 (Chapman 1973) with an estimated fixed pregnancy rate of 60%. Since 1960 the size and yield of the herd has decreased leaving Chapman (1981) to conclude that the predictions of the original models have not been fulfilled. The early success of the Russian overseers resulted from their insight into fur seal biology which enabled them to separate the commercial harvest from the reproductive potential of the herd. American biologists were also successful at applying the same simple model to the fur seal population until the early 1950s when they tried refining it. Since this time, concentrated research efforts have produced a plethora of information about the northern fur seal. But there is little sign that any of it has improved the original Russian model of population management.  1.2  Thesis Overview  The Russian fur merchants recognized early on the need to understand the basic biology of the seal. I try to build upon this base by exploring questions concerning why fur seals breed on the Pribilof Islands and the obstacles they have successfully overcome to do so. I also consider the timing of migration and timing of birth in relation to climatic conditions and food availability. Data are presented describing the growth of males and females from the point of conception until adulthood. Thus I attempt to combine basic biological relationships and put the fur seal into an evolutionary and ecological context rather than focusing solely upon it as a species isolated from its environment. A major theme throughout the thesis is population management.  I highlight  changes that occurred in the fur seal population as densities changed and consider how this information might be applied in a management context. I am particularly  Chapter 1  General Introduction  7  interested in why the American refinement of the Soviet model for population management failed and whether or not the original simple model can be improved. I also address the question of whether density dependent processes can be monitored and used to manage pinniped populations. I rely primarily on two data bases. The first is from land based studies spanning the period 1911-89 (Lander 1980a). In a few cases some summaries of the land data have been previously published, while in others the data were key punched but not thoroughly analyzed or were discovered in field notebooks in the fur seal archives of the National Marine Mammal Laboratory, Seattle WA. The second major data set was collected at sea from 1958 to 1974 by the Canadian Department of Fisheries and Oceans and the United States Marine Fisheries Service (Lander 1980b). Preliminary analysis of this data set was completed in the late 1970s, but only the fur seal diet was thoroughly studied. In addition to these two data sets I also draw upon the results of behavioural observations and physiological studies conducted in both the field and laboratory on northern fur seals and other pinniped species. I try throughout to be rigorous in my selection of data and am cautious to avoid possible biases that can result from negligent pooling of data across time and place without regards for the effects of seasonal and locational errors. While this may seem tedious at times, I believe it is necessary to ensure that incorrect conclusions are not spuriously drawn from the data. I also devote considerable attention to establishing and understanding basic biological relationships that give rise to the data before attempting to interpret interannual changes that might have occurred in the population.  1.3  Survey of Individual Chapters  Each chapter is written as a paper that can stand on its own. As such there is some necessary overlap in the material presented in a few of the chapters. In particular  Chapter 1  General Introduction  8  the reader may wish to skip over review material that appears in the first few pages of each chapter. The general organization of the chapters follows the development of the fur seal from conception and birth through to sexual maturity and finally to a synthesis of the earlier material in terms of population regulation and management, and reasons for the decline of the Pribilof herd. The first two chapters consider fetal growth from 1958-74. Chapter 2 is an in depth analysis of sex specific growth patterns and the effects of maternal size and age on fetal growth. The lessons learned from this analysis are applied in Chapter 3 to estimate the annual size of fur seals at birth and determine whether changes in fetal growth can be related to the build up of commercial fishing that began in the early 1960s. In addition, a condition index describing the relationship between body mass and length is proposed to compare the effects of biotic and abiotic factors on the well-being of individual seals and the population in general. Chapters 5 to 9 focus upon the growth and survival of newborn fur seals. It became necessary in the course of this work to understand the timing of birth and establish the mean date of birth to assign ages to the pups collected over the summer and to identify the time period when pups would be most vulnerable to inclement environmental conditions. Thus I developed two methods in Chapter 4 for estimating the mean date of birth from daily counts of live pups. The ability of fur seal pups to cope with diverse climatic conditions on land was investigated by constructing a thermal budget based on published physiological studies (Chapter 5). This represents the first attempt to gain a holistic understanding of neonatal fur seal energetics and its environmental implications. The precision of the timing of birth (Chapter 4) and impact of climate on pup survival (Chapter 5) suggest that weather conditions might be one of the driving forces behind the fur seal's life cycle. This is further investigated in Chapter 6 using hourly weather records recorded on St. Paul Island from 1956-86.  Chapter 1  General Introduction  9  The next three chapters deal with pup weights. First the question of whether tagging and handling affects subsequent growth is addressed (Chapter 7). The results of this chapter are important for interpreting the results of subsequent analysis that are based on measurements from tagged individuals. Next I assess how milk consumption, the timing of birth, and the effects of growth and sample size can influence the size of pups captured for weighing (Chapter 8). The results show hidden biases related to human error and fur seal biology that must be controlled for when weighing northern fur seals and other fur seal species. Finally, changes in the size of pups weighed from 1957 to 1987 as the population declined are assessed and related to fluctuations in the Bering Sea food supply (Chapter 9). Chapters 10 and 11 are concerned with the growth of males and females from weaning to maturity. The first chapter shows that seasonal increases and decreases in body mass and length occur, and that seals do not grow continuously throughout the year. I also relate body growth to the timing of migration and show the importance of fur seal feeding areas outside of the Bering Sea. Chapter 11 builds upon the premises of fur seal growth and tests the hypothesis that body size increased as population density decreased. I also present evidence that body growth might be influenced by large scale environmental factors. The next two chapters are related to the male harvest on St. Paul Island. Chapter 12 estimates the ability of sealers to capture males at haulout sites and describes the age composition of animals on land. The data were pulled from a cardboard box by J. Scordino (NMFS, Seattle WA) after I spoke with him about estimating juvenile survival rates using, among other parameters, an estimate of capture efficiency. From his field notes, I assess the factors that influence haulout composition and estimate escapement rates during the annual harvest. No other study of this kind has been done before. I discovered another data set pertaining to the male harvest in the filing cabinets of the fur seal archives (NMML, NMFS Seattle, WA). Hand  Chapter 1  General Introduction  10  written and mimeographed sheets contained the daily measured lengths of over 1.6 million seals killed between 1933 and 1956. This data set is combined in Chapter 13 with the lengths of known-age males killed from 1943 to 1983 to determine what changes occurred in body size as the population density rose through the 1930s and 40s and fell from 1956 to present. Chapter 14 estimates annual pregnancy rates and age at sexual maturity during the years when females were harvested and population density dropped (1956-68). The results of this study and estimates of survival rates and changes in morphometric measures noted in earlier chapters are synthesized in Chapter 15 which examines whether density dependent changes in pinnipeds are useful for predicting future population trends and assessing current population status. The physiological conditions and vital rates that have apparently responded to changes in population density are identified and the significance of such changes for the management of pinnipeds in general is discussed. Finally Chapter 16 relies on information presented in previous chapters to examine why the Pribilof fur seal is declining. Two hypotheses are reviewed and a new, third hypothesis is proposed. The chapter also outlines some directions for future research.  Chapter 2  Fetal Growth: Life History Strategy and Sources of Variation 2.1  Introduction  Quantitative descriptions of fetal growth are useful for exploring the evolution of various life history strategies (e.g. Huggett and Widdas 1951; Stephenson 1962; Evans and Sacks 1973; Frazer and Huggett 1974; Frazer 1977; Case 1978; Clader 1982; Peters 1983). They axe also potentially useful when applied on a population level to assess environmental stresses, or used to interpret the impact of maternal age, size and parity on the developing fetus. This is particularly relevant to studies that make inferences about population status from the size of individuals measured at birth. There are few studies of fetal growth in pinnipeds (see Hewer and Backhouse 1968; Stewart et al. 1989). This is due to the large number of samples required to accurately quantify growth and the difficulty of obtaining them. Early studies of northern fur seals  (Callorhinus ursinus) provided anatomical descriptions of the  developing fetus and discussed the timing of implantation (Enders et al. 1946; Pearson and Enders 1951; Baker 1957; Scheffer 1960, 1962; Craig 1964, 1966), but were  11  Chapter 2  Fetal Growth  12  insufficient for quantifying the fetal growth process because of limited sample sizes. In 1958 a pelagic research program was started by Canada and the United States as members of the North Pacific Fur Seal Commission (Lander 1980a). By 1974 over 7 000 fetuses had been collected. Preliminary analyses of this data set indicated that males fetuses were larger than female fetuses (Fiscus et al. 1964; Lander 1979a; York 1987). It was also suggested that the size of fetuses increased with the age of the mothers (Lander 1979a). In the current study I present an in-depth analysis of the fetal data collected from 1958-74. I begin by briefly reviewing the reproductive biology of northern fur seals, and proceed with an analysis and discussion of the biases contained within the data set. My aim is to test for differences between the size of male and female fetuses and to describe growth curves for the length and mass of each sex. I also test whether there is any difference between the size of fetuses taken from primiparous and multiparous females, and whether the age and size of the mother has an effect on fetal growth.  2.2  Biology of Fur Seal Reproduction  Pregnant northern fur seals return to land, in order of decreasing age, to give birth to a single pup (M.A. Bigg pers. comm.). The earliest births on the Pribilof Islands occur shortly after the first females arrive on land on about June 20 (Bartholomew and Hoel 1953; Peterson 1968). Few pups are born after July 20. Over 50% of the females arrive to give birth during the first two weeks of July (Chapter 4). This pattern of birth appears to occur at the same time each year (Gentry and Francis 1981; Bigg 1986; Chapter 4). Direct observations indicate mating occurs 3 to 5 days after parturition (Bartholomew and Hoel 1953, Dorofeev 1961), and ovulation follows coitus (Craig 1966). For 3.5 to 4 months after fertilization, the blastocyst is free in the uterus, bathed  Chapter 2  Fetal Growth  13  in uterine secretions. The blastocyst moves to one of two uterine horns in early to mid November (Craig 1966; Yoshida et al. 1978). From there it passes through the epithelial layer of the endometrium and begins to form tissues and organs. The period of delayed implantation from mid July to mid November coincides with the lactation period (Craig 1966; Daniel 1981). A mean photoperiod of 12.5 h d  _ 1  occurring 62 days after the mean date of parturition (or 65 days before implantation) may act as a cue for the initiation of implantation (Temte 1985).  2.3  Materials and Methods  Canadian and American biologists collected reproductive tracts and morphometric measurements from female fur seals along the coasts of Canada (British Columbia) and the United States (California, Oregon, Washington, and Alaska) from 1958 to 1974 (Lander 1980a). They sectioned the ovaries and recorded the condition and diameter of follicles, the formation or presence of a corpus luteum, and the presence of corpora albicantia. On the basis of these examinations each female was classified as nulliparous (female has never been impregnated), primiparous (first time impregnated, regardless of whether the pregnancy is carried to term) and multiparous (previously pregnant and either pregnant or nonpregnant at the time of collection). The dead fetus was sexed, weighed and measured (crown - rump length) if normal, and otherwise noted as aborted or reabsorbed. Adult females were weighed (with their fetus), measured (tip of nose to tip of tail, belly up, on a measuring board), and aged (by counting dentinal annuli in canine teeth). Birthdays had been assigned to January 1 (Lander 1980a) rather than July 1 when they were born. The age data were therefore adjusted to reflect the true biological ages of the samples as discussed by Trites and Larkin (1989). The timing of collection during leap years and non-leap years was standardized by assigning Julian dates.  Chapter 2  Fetal Growth  14  Exploratory data analysis was conducted with S (Becker and Chambers 1984), and statistical analyses were completed using BMDP software (BMDP 1988). 2.3.1  Verification and Detection of Biases in D a t a Collection  Data were plotted and examined for outliers. These were verified by comparison with the original field notes. Possible biases in the sexing of fetuses and in the over- or under-sampling of primiparous females by month were tested by contingency analysis. The proportion of male fetuses and the proportion of primiparous females captured were determined by month, and confidence limits for the proportions calculated (Zar 1984). Similarity in the maternal age distribution of the samples between months and years was visually inspected using boxplots and statistically tested by analysis of variance.  Post hoc pairwise comparisons identified which group differences ac-  counted for significant overall F values. The equality of group variability was tested with Levene's test, while Box-Cox diagnostic plots (log of standard deviations versus log of the group means) assisted in selecting transformations to achieve greater homogeneity of variance. Welch (1938) and Brown-Forsythe (1974) procedures were used when the homogeneity of variance assumption was not met. 2.3.2  Average Fetal G r o w t h  . Changes in the body length of male and female fetuses were described by four sigmoid growth curve models for the purpose of comparison: the Richards (Richards 1959), von Bertalanffy (von Bertalanffy 1938, 1957, 1960), Gompertz (Gompertz 1825), and logistic (Verhulst 1838) models were fitted using a nonlinear least squares procedure. Nonlinear methods have been shown to have some superior structural properties that produce more accurate and precise parameter estimates than linear methods (Vaughan and Kanciruk 1982).  Fetal Growth  Chapter 2  15  Growth of body mass (M) was linearized and estimated as a function of gestational age (r). In theory, the dimension of mass varies with length cubed (L ) 3  and time cubed (t ), assuming geometric similarity (Gunther 1975). The actual 3  relationship between fetal length and mass was estimated by linearizing the relationship M = aL  b  such that log(M) = log(a) + b log(D) where b is the slope and  log(a) is the intercept. It has been suggested that the geometric mean of the regression of mass on length and the inverse of the regression of length on mass should be used to give the appropriate linear regression for mass-length comparisons because length is not truly independent of mass (Ricker 1973, 1975, 1979). However, others have indicated that ordinary least squares regression is appropriate and easier to interpret than geometric mean regression (Sprent and Dolby 1980; Cone 1989). Thus I conducted ordinary least squares regressions. Separate growth curves were determined for 3 239 male and 3 377 female fetuses. Data from parous females collected from November to June were combined, but samples collected in July were excluded because they were not considered random. 2.3.3  Differences Between Primiparous and M u l t i p a r o u s Females  Differences between the sizes of fetuses carried by primiparous and multiparous females were tested by analysis of covariance, correcting for the gestational age (i.e. Julian date when sampled) and the age and size of the mother. Possible biases in data collection were reduced by restricting the analysis to females aged 4 - 7 y that were sampled from February to June, a period of time when fetal growth can be linearized. Mass to the power of 0.36 and the square root of gestational age (measured in Julian days) were the only transformations required to linearize the relationship between the dependent variable and the covariates (Chapter 3). A parallelism test indicated whether the slope between the dependent variable and the  Chapter 2  Fetal Growth  16  covariate was parallel for all groups (BMDP 1988). Differences between the size (length and mass) of primiparous and multiparous females were also tested by analysis of covariance after adjusting for the effects of age and annual growth. The mass of the mother was set equal to her total mass minus that of her fetus. The logarithm of the resulting mass and the square root of gestational age (Julian day) linearized the relationship between these variables. 2.3.4  Effect of M a t e r n a l Age on Fetal Size  Much of the variability in the size of adults and fetuses collected over a year is due to differing stages of growth and development when sampled. As in the previous analysis, this can be removed by using the day of capture as a covariate. In this way, an analysis of covariance tested whether the fetal size of multiparous mothers was dependent upon the mother's age. The dependence of fetal size (length and mass) on maternal age was further described using a quadratic regression. If the quadratic term is negative and contributes significantly to the regression, it can be concluded that older females have smaller fetuses. A t-value equal to the ratio of the coefficient to its standard error was used as a two tailed test of significance for this quadratic coefficient. The analysis was restricted to multiparous females collected from February to June, because of possible biases in fetal size related to parity.  2.4 2.4.1  Results and Discussion Verification of D a t a  Screening the length and mass data for outliers and distributional consistency revealed over 300 possible errors in measurement or data entry. Consistency between the length/mass relationship confirmed the reliability of about half of these suspect observations. A further 20 were key punching errors and were corrected. The remaining suspect data were rejected. In one case the lengths of the 78 fetuses col-  Chapter 2  Fetal Growth  17  lected during June of 1959 exceeded expected lengths when compared to the known length/mass relationship and when compared to the expected lengths based on other June samples as well as those collected earlier in the year. It appears in this case that the observer may have measured curvilinear length rather than crown-rump length. In other cases the measurements may have been incorrectly written on the data card. Inspection of the original data cards further suggests that some of the errors might have arisen when several animals were on the ship deck at one time, and the measures inadvertently recorded on the wrong card. Despite these shortcomings, the number of suspect measurements rejected is few (202) and the number of remaining observations considerably large (7 003). Over 60% of the fetuses measured were collected during the first four years of pelagic surveys, 1958-61 (Table 2.1). During this period, approximately 1100 fetuses were collected annually from a population of approximately 350 000 pregnant females (Lander 1980b; Trites 1989). From 1962 to 1972 the number of fetuses measured was reduced to an average of 240, then to under 40 in 1973-74. Pooled fetal samples (1958-74) cover the entire period of development from implantation to birth (Fig. 2.1). The only gap in data collection was during a two week period at the end of December, beginning of January when the biologists and ship's crew were ashore on holiday. It is apparent from Fig. 2.1 that growth of northern fur seal fetuses is synchronous. Few of the fetuses failed to conform to the general pattern of growth. None of the fetuses collected were exceptionally large compared to the others, but some (< 50) were unusually small (see Fig. 2.1). These small fetuses (< 0.7% of all fetuses sampled) were proportionally the correct size (based on their length-mass relationship) but had obviously implanted much later than the rest of the population. With the exception of very small fetuses, the distribution of fetal lengths and  Chapter 2  Fetal Growth  18  Figure 2.1: Length and mass of 7 003 fetuses by date of sampling. A small amount of random noise was added to each variable to reduce the overlap caused by measurement roundoff. The letters on the date axis denote the beginning of each month.  Fetal Growth  Chapter 2  19  Table 2.1: Number of fetuses collected at sea by month from 1958-74. No fetuses were sampled in September and October. Of the 7 003 fetuses collected, 98 were not sexed and 156 were either not measured or not weighed.  Year 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1958-74  Jan 0 95 0 311 0 0 0 0 30 41 55 0 32 26 19 22 25 656  Feb 154 576 1 284 32 0 0 0 105 26 93 51 18 2 2 8 0 1352  Mar 337 202 165 100 3 0 0 0 126 0 6 87 66 87 38 0 0 1217  Apr 143 324 290 204 14 0 58 86 • 21 16 51 40 18 48 23 0 0 1336  May 391 34 364 26 20 12 108 61 3 5 75 32 54 28 47 0 0 1260  Jun 95 0 150 1 329 112 6 35 0 0 137 0 2 0 0 0 0 867  Jul 0 0 38 0 65 67 22 0 0 0 21 0 0 0 0 2 15 230  Aug 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1  Nov 0 0 0 0 0 0 0 0 0 3 0 0 0 0 0 0 0 3  Dec 0 0 2 0 0 0 0 0 0 63 0 0 0 11 2 3 0 81  Jan-Dec 1120 1231 1010 926 464 191 194 182 285 154 438 210 190 202 131 35 40 7003  masses do not indicate any substantial problems with measurement error (Figs. 2.1 and 2.2). But Fig. 2.2 suggests the mass of small fetuses may not be overly reliable. Perhaps this is due to the difficulty of separating the tiny fetus from uterine tissues and fluids. The large variation in fetal mass at this stage of development may also be related to the difficulties of weighing aboard ship. 2.4.2  A g e and P a r i t y Biases  The age distribution of the pregnant mothers was positively skewed (Fig. 2.3). Most females began reproducing at the age of 4 or 5 years, although a few were as young as 3 years old and one was only 2 years old (c/. Chapter 14). The oldest primiparous female was 15 years old, but this and other observations of old primiparous females are highly suspect.  Pregnant females were dassified as primiparous if a corpus  Fetal Growth  Chapter 2  20  2 CO CO  O  1 0 -1  ^  -  |  -3  ^  -4 -  2  -5 -  i—i—i—r 1—i—i—i—i—i—r -  N D J  F M A M J  J A S  Date (d) Figure 2.2: Natural logarithm of fetal mass by date of sampling. A small amount of random noise was added to each of the 7 003 data points to reduce the overlap caused by measurement roundoff. The letters on the date axis denote the beginning of each month.  albicans was not present. A corpus albicans was presumed to indicate a previous pregnancy, although in actual fact it results from ovulation regardless of whether pregnancy ensues. Furthermore the absence of a corpus albicans does not necessarily mean the seal is primiparous because corpora albicantia degenerate over time and may not be present 2-4 years following a pregnancy (Craig 1966). Thus many of the older females recorded as primiparous (> 8 y ?) were probably misclassified. The relative frequencies of primiparous and multiparous females collected from December to July differed by month (Pearson X7 = 35.48, p < 0.001). Proportionally more primiparous females were sampled during December and July than at any other time of the year (Fig. 2.4a). But no significant difference could be detected in the relative frequencies of primiparous and multiparous females taken from January to June (Pearson x l = 5.73, p < 0.33). The age distribution of pregnant females collected was similar in all months, but three: January, April and July (Fig. 2.4b). The seals sampled during these three months were, on average, younger than at other times of the year. The significance of  Chapter 2  Fetal Growth  21  Age (y) Figure 2.3: Number of primiparous and multiparous females captured by age from 195874. these monthly differences was assessed by analysis of variance using the logarithm of age to reduce the skewness. But the transformation failed to significantly equalize variances (Levene's ^7,6905 = 4.01, p < 0.001). Therefore the Welch (^7,1019 = 7.20, p < 0.001) and Brown-Forsythe ( . F 7  1 9 3 8  = 7.55, p < 0.001) tests were used.  They confirm the bias in ages sampled from December to July. The under-representation of older animals in April was not expected and cannot be readily explained. However, the over-sampling of young and primiparous fur seals in December and July probably reflects the timing of migration to and from the breeding islands. Females appear to return to the Pribilof Islands in order of decreasing age and give birth within a few hours or days of arriving on land (Bartholomew and Hoel 1953; Peterson 1965, 1968; Bigg 1984, 1986). Pupping begins about June 20 andfinishesJuly 20 (Bartholomew and Hoel 1953; Peterson 1968; Chapter 4). Therefore many of the older - multiparous females have already given birth by the end of June, and are not available for sampling in July. The females suckle their pups until October, early November, whereupon they leave the Pribilofs and begin their southward migration (Peterson 1968). Those seals impregnated for the first time are not encumbered with a pup and presumably  Fetal Growth  Chapter 2  22  30 -  V) O h—  CO  g. 'E  •=  0.  20 -  0  10 0  0  0  (D Q  (t)  !  i  i  i  i  hr  •  20  S  i  1 5  <  i  5 0  -  55 co  50  re  45 H  -1-  C)  fh  0 <\  db db  i  35  M  A  M  Month  Figure 2.4: Biases in monthly samples. A . Percentage of pregnant females each month that were primiparous. The size of the circles surrounding each estimate is proportional to the sample size and the bars are 95% confidence intervals. The solid line is the mean estimate for all months combined. B. Age distribution of the pregnant females collected. The line in the middle of each box shows the median. The central 50% of the data is shown by the length of the rectangle and the dashed lines show how stretched the tails of the distribution are (25% of the data in each tail). The width of each box is proportional to the sample size and outliers are shown with dots. C . The sex ratio of fetuses collected. The solid line indicates the expected 50:50 ratio. Again the size of the circles surrounding each estimate is proportional to the size of the sample and the bars are 95% confidence intervals. The total sample size was 6905 seals.  Fetal Growth  Chapter 2  23  leave the islands earlier. Thus the age samples collected in December in the north Pacific are probably not well mixed and reflect the segregated fur seal departure. The Welch ( W  16i891  = 11.60, p < 0.001) and Brown-Forsythe tests (Ff  6<2274  =  10.41, p < 0.001) indicate a highly significant difference in the mean age of seals sampled each year during 1958-74. This result is not surprising. Age biases could have arisen because the migrating seals were not well mixed and the months sampled differed between years (Table 2.1). The bias might also reflect differences in year class strength caused by high variability in annual mortality rates of young animals (Trites 1989). Similarly, the annual age structure was undoubtedly altered by the harvesting of 315 000 adult females that occurred from 1956-68 (Lander 1980b). 2.4.3  Sex R a t i o Biases  The relative frequencies of male and female fetuses (i.e. the sex ratio) collected each month were the same from February to July (Pearson xl = 4.67, p = 0.59), but differed with the inclusion of the December and January samples (Pearson xl — 14.87, p = 0.04; see Fig. 2.4c).  Significantly more female than male fe-  tuses were collected in December and January than any other month. The most likely explanation for this result is human error in sexing the small developing fetus. If fetuses were sexed by the presence or absence of a penis, there is a higher probability of misclassifying a fetus as female because the penis is not always clearly distinguishable. It also takes some time for the fetus to develop to the point that gender can be identified. This means that all of the identified males were likely true males, but that some of the recorded females were in fact males. It has been suggested that sex ratios of the fetuses might be biased by the physical condition of the mother (Trivers and Willard 1973). If so, young mothers might produce more females than older mothers because the smaller size of female fetuses would require less maternal resources (e.g. Thomas et al. 1989). However, there is no  Chapter 2  Fetal Growth  24  indication of such a phenomenon occurring in northern fur seals. The frequency of males and female fetuses carried by mothers of different ages (grouped as in Fig. 2.5) from February to July did not differ from unity (Pearson x%  =  5.08, p — 0.53).  Similarly there was no difference in the sex ratio of fetuses carried by primiparous mothers (x\ — 1-40, p — 0.50) or by multiparous mothers (x% = 5.71, p = 0.46). Furthermore, there was no difference in the sex ratio of fetuses collected from one year to the next from 1958-72 (x\ = 14.18, p = 0.44). In all, 3067 male and 3 097 A  female fetuses were collected from February to July. 2.4.4  Average Fetal G r o w t h  Male fetuses are bigger than female fetuses (Fig. 2.6). Coefficients of linear regressions of male and female lengths by age and female masses by age  (i*2,6884  (^2,6906  = 413.71,  = 273.73,  p < 0.001) and male  p < 0.001) are both significantly dif-  ferent. The growth differential between the sexes begins shortly after implantation and increases through to birth. Sigmoid growth curves were reasonable models for describing changes in body length. The von Bertalanffy model proved to be marginally better than the Gompertz and Logistic models (Table 2.2). But parameter estimates for the Richards equation failed to converge. This might be explained by the number of data points fitted or by the efficiency of the computer algorithm used (Fitzhugh 1976). Some authors have further reported difficulty in fitting the Richards equation because of the high correlation between 2 of the 4 parameters, K and n (Rutledge et al. 1972; Brown et al. 1976; Davies and Ku 1977) Several authors have encouraged the use of the Richards equation because it is the general form of the other growth equations and permits comparisons with growth curves constructed for other species (see Brisbin et al. 1987; White and Brisbin 1980).  However the Richards model is effectively a 3 parameter model  Fetal Growth  Chapter 2  25  150 -g-140 ^ 1 3 0 ^  0)120  c  110 100 80  E o_ ra _  CD  B  6 0  BE^B  40 20 0 8  4  2  -  0  -  —, <5  1  6-7  1  8-9  1  10-11  1  12-13  1—i—  14-15  16+  A g e (y)  Figure 2.5: Maternal size and the effect of maternal age on fetal size in multiparous females. A . The distribution of maternal lengths of multiparous females by age. B . &c C . The effect of maternal age on fetal length and mass in multiparous females. In all, the distributions include 5276 fetuses collected during the entire sampling season from early implantation in December to full term at the end of June. The notches on the boxplots provide an approximate 95% test of the null hypothesis that the true medians are equal. If the two notches overlap then the null hypothesis is not rejected with (approximate) 95% confidence. The width of each box is proportional to the size of the sample.  Fetal Growth  Chapter 2  80  E  ,CJ  26  -  60  • £ 40 CO _§  20 0 8  o5  6  4  -  2  -  0  -  i I I I I I I r  N D J F M A M J J A S  Date (d)  T  1  r  20  40  60  T 80  Length (cm)  Figure 2.6: Growth curves for male (solid line) and female (dashed line) fetuses. Length is described by the von Bertalanffy equation (Table 2.2). Equations describing changes in body mass and the length-mass relationship are contained in the text. Confidence limits (95%) were calculated but are indistinguishable at the current scale from the curves drawn. The letters on the date axis denote the start of each month.  because 2 of the 4 parameters are highly correlated (Zach 1988). Furthermore, the Richards model usually fails to explain a larger portion of the variation in the data than the simpler 3 parameter models (Zach et al. 1984). Given these limitations and the performance of the Richards model when fitted to the fetal data, its value is questionable. Unlike length, fetal mass does not approach an asymptote. Instead, body mass appears to increase exponentially, never reaching an inflection point (or perhaps reaching it at parturition). Human fetal studies further support the observation that the initial increment of body mass is slow, followed by rapid exponential growth to a peak value near term (Vorherr 1975). Growth (mass) of human fetuses slows  Fetal Growth  Chapter 2  27  Table 2.2: Three growth models representing the length of fetuses (L ) in cm over time t recorded in days (where t = 0 represents Nov. 10). The data, 3 239 male and 3377 female fetuses, were pooled from November to June for the years 1958-74. Parameters A, c, and K are the asymptote, the constant of integration, and the growth rate constant, respectively, as defined by Zach et al. (1984). t  Model  Equation  von Bertalanffy  Lt =  Gompertz  L =  Logistic  L = A/(l + ce- )  t  A{l-ce- f Kt  Ae- ~  t  Parameters c 84.55 .76 77.91 .79 3.34 78.16 72.54 3.49 69.30 10.69 65.00 11.38  A  ce  Kt  Kt  K  Residual Sex Mean Sq.  .009 .010 .012 .013 .020 .021  10.45 10.38 10.53 10.47 10.92 10.95  m f m f m f  markedly after the normal gestation time of 38 weeks, and stops after 40 weeks. Thus the apparent asymptote in fur seal fetal mass that abruptly occurs in July (Fig. 2.1) may be indicative of prolonged pregnancies rather than the result of over sampling fetuses from the small late arriving primiparous females. Changes in male and female mass (M kg) with gestational age (t days), can be described for t > 35 by the equations  M = (-1.214 + .207 y/i) -  2 75  and  M =  (—1.183 + .201 y/i) - , respectively, where t = 0 represents November 10 and t = 50 2 75  is January 1 (Fig. 2.6). November 10 is approximately the time that the blastocyst moves from the uterus to one of two uterine horns (Craig 1966). A plot of the residuals against fitted values shows that the model under-estimates mass during December and January. This suggests that two distinct stages in the growth of body mass may be present. The models provide good representations of increases in tissue mass from February to June, but do not adequately describe the initial two month stage. This result is in keeping with fetal studies of other species that suggest there is an initial lag phase of development during which the placenta becomes established (Payne and Wheeler 1967a).  Fetal Growth  Chapter 2  28  The relationship between fetal length (cm) and mass (kg) for both sexes combined was M = 1 0 - X ' . Note that the exponent is significantly less than the -4  20  2  75  predicted value of 3.0 (t&s23 = 47.16, p < 0.001). Individually, the mass-length relationship was  M = 10~ - L 4 23  277  for male fetuses and  M = 10~ L 418  274  for females.  While this difference between the exponents of the male and female mass equations was small, it was nevertheless statistically significant  (^,6821  = 86.33, p < 0.001).  It means that a male fetus that is the same length as a female fetus will weigh marginally more. Males weigh more than females because their skeleton (measured as length) grows faster than the female's. Greater mass can presumably be accumulated on the larger male frame. The most rapid increase in fetal length occurs during the first half of pregnancy, but body mass does not increase rapidly until the latter part of pregnancy. The possible bias detected earlier in sexing the fetuses during December and January, does not affect the growth patterns described for males, but may slightly increase the pooled mean size of females during the first two months because a few of the males were incorrectly sexed as females. However, the number of individuals is small, as is their size, and would not significantly alter the results. The July data were excluded from the analysis because young and primiparous females were overrepresented in this month. If young and primiparous females have smaller fetuses than other females, including the July data would have underestimated fetal growth. 2.4.5  Differences Between Primiparous and M u l t i p a r o u s Females  There was a significant difference between the adjusted mean size of primiparous and multiparous females after the effect of the mother's age and the sampling date had been removed. Primiparous females were found to be lighter  (//i,2054  = 15.59, p <  0.001) and shorter (.Fi,2064 = 13.36, p < 0.001) than multiparous females. Similarly  Chapter 2  Fetal Growth  29  primiparous females carried smaller fetuses than multiparous females. Differences in fetal length (Ji, 052 = 54.80, p < 0.001) and mass (f 052 = 74.72, p < 0.001) 2  lf2  are attributable to the parity of the mother. However, the reason that multiparous females have bigger fetuses cannot be entirely explained by their larger maternal size. It appears instead that females are somehow physiologically altered by their first pregnancy. Perhaps, for example, increases in the numbers of uterine blood vessels associated with the growth of an earlier fetus, provide better nutrition to developing fetuses in successive pregnancies. All of the assumptions of the analysis of covariance were met. In all cases, there was a linear relationship between the dependent variables and the covariates in each group. Furthermore the p values for the test of equality of slopes were all nonsignificant. 2.4.6  Effect of M a t e r n a l A g e on Fetal Size  Fetal size is dependent upon the age of the multiparous mother.  There were  significant differences among the mean size of fetuses carried by multiparous females of different ages after the effect of gestational age (Julian date) was removed (length: i^,4368 = 39.64, p < 0.001; mass: F ,4355 = 48.26, p < 0.001). Older females 4  had increasingly bigger fetuses until reaching their apparent reproductive prime, based on fetal size, at the age of 10-11 y (Fig. 2.5). As females continued to age, their fetuses became increasingly smaller. This 'senility' effect is presumed real, based on the significance of the quadratic coefficient of the polynomial regression (length: <608i = -3.26, p < 0.01; mass:< 063 = -3.02, p < 0.01). 6  Old females may produce small fetuses, but they themselves are in fact very large (Fig. 2.5a). While the annual growth increment of adult females decreases with time, they nevertheless grow continuously and may never reach an asymptote before dying (Fig. 2.7). This result is surprising, but is consistent with the fact that  Fetal Growth  Chapter 2  30  J  130 128 -  E  •H- 126  -£Z  "jf  CD _J  -  ,4"  124 -  ,4  122 120 -  /  T — i — i — i — i — i — r <S  6-7  10-11  8-9  12-13  14-15  16-t  Age (y) Figure 2.7: Mean length of multiparous females collected from December to June (Fig. 2.5A). The vertical bars are 95% confidence limits on the means. The growth curve was fit by linear regression.  annual growth increments are regularly noted on hard body parts such as teeth. It is also supported by the observation that many species of fish, amphibians and reptiles, which live in the supporting medium of water, continue to grow beyond the self retarding plateau phase (Batt 1980).  2.5  Overview  Most pinnipeds are sexually dimorphic, with males being larger than females (Alexander  et al. 1979; King 1983). Data from grey seals, Halichoerus grypus (Kovacs  and Lavigne 1986; Anderson and Fedak 1987), northern elephant seals,  Mirounga  angustirostris (Le Boeuf et al. 1989), and southern elephant seals, Mirounga leonina (McCann et al. 1989) show that the mass of males exceeds that of females at birth. Among otariids such as the Antarctic fur seal,  Arctocephalus gazella (Doidge et al.  1984a), northern fur seal (Costa and Gentry 1986), Galapagos fur seal  Arctocephalus  galapagoensis (Trillmich 1986), and California sea lion Zalophus californianus (Oftedal et al. 1987), it has been shown that male pups are bigger at birth and proceed to grow faster than female pups.  Chapter 2  2.5.1  Fetal Growth  31  Observed and Predicted Neonatal Size  Based on the fetal growth curves (Fig. 2.6), male northern fur seals are 10% heavier and 5% longer than females at birth. On June 30, the average male pup is predicted to be 63.6 cm long and weigh 6.1 kg. Females should be 60.5 cm long and weigh 5.6 kg. Measurements of newborn pups taken on the rookeries indicate a mean mass of 5.7 kg for males and 5.1 kg for females (Scheffer and Wilke 1953; Lander 1979a; Fowler 1990a). The average length for both sexes was about 62 cm. Comparing these field estimates with the predictions of the fetal growth curves suggests that newborn pups maintain their length but lose about 8% of their body mass after birth. The discrepancy between observed and predicted body mass at birth may be real or may be an artifact of sampling biases. For example, in a study of southern elephant seals, McCann et al. (1989) weighed neonates and noted that growth (measured as mass) was slow or negative for the first 2-3 days for the majority of pups. There is also evidence that human infants usually lose some mass in the first 2 to 5 days after parturition, possibly as much as 10% of the infants birth mass (Reeder et al. 1983). The mass loss at birth may be related to the hormone, estrogen which increases water retention and blood volume, and is likely produced at high levels by the female fur seal during pregnancy. The pup would be born with high levels of the mother's estrogen. But after birth, the pup's ability to retain water is reduced because the source of estrogen is removed, and the pup loses body mass for a few days. Another explanation for the discrepancy in neonatal mass of pinnipeds is related to the timings of birth and field work. If field work is conducted in mid July, most of the newborns will have young primiparous mothers, and will hence be smaller than pups born earlier in the season. This has been shown to be the case for Antarctic  Chapter 2  Fetal Growth  32  far seals. Boyd and McCann (1989) found that smaller adult females returned over the 40 day breeding season, and that they produced progressively smaller pups. In fact, the birth mass of male and female Antarctic fur seal pups (which are about the same size as northern fur seal pups) were respectively about 0.5 kg and 0.4 kg lighter at the end of the breeding season than at the beginning. 2.5.2  L e n g t h of Gestation  Little is known about the gestation time of primiparous and multiparous females. Since primiparous females pup later in the breeding season than multiparous females it is possible that they also implant later in the season. However it has been suggested that photoperiod, through an effect on pineal gland secretion, is an exogenous cue to initiate implantation and later synchronization of the timing of parturition (Keyes et al. 1971; Elden et al. 1971; Temte 1985). Thus all females should implant at about the same time. Younger and primiparous females would have longer gestation times if implantation is synchronous because they give birth after the older multiparous females. Extending the gestation period gives the fetus more time to grow and would presumably result in producing a larger pup than normal at birth. However, human studies reveal that prolonged pregnancies result in degeneration and calcification of the placenta which causes fetuses to lose mass and increases their postnatal mortality rates (Vorherr 1975; Knuppel and Drukker 1986). There is also evidence of higher fetal perinatal mortality for primiparous humans than multiparous females, particularly when the pregnancy proceeds beyond term (Vorherr 1975). A further disadvantage in prolonging the birth of northern fur seals is that the neonatal pup will have less time to suckle while on the rookery before the breeding season ends. This may have bearing on the observation that neonatal mortality among Antarctic fur seals increases as the breeding season progresses and increasing numbers of  Chapter 2  Fetal Growth  33  pimiparous females give birth (Doidge et al. 1984b). 2.5.3  M a t e r n a l Resources  Boyd and McCann (1989) suggest that growth of the male Antarctic fur seal fetus is limited by maternal size and resources, but that female fetuses do not fully exploit their mother's resources. This could well be the case for northern fur seals as well. Note however that another study of Antarctic fur seals (Costa et al. 1989) and one on southern elephant seals (McCann et al. 1989) found that female pups were more closely related to maternal size than were males. Unfortunately, this observation may be inconclusive because these two studies did not correct for the change in maternal mass and neonatal size that occurs through the pupping season. An estimate of the condition of pregnant northern fur seals found that females carrying male fetuses were in poorer condition than those with female fetuses (Chapter 3). Male fetuses grow faster and demand more resources than female fetuses. Even when fetal length is standardized for both sexes, males are marginally heavier. Thus it seems that males are the 'greedy' sex by attempting to grow as large as possible. It is curious that female fetuses do not fully use the maternal resources, given that a larger birth size would probably infer a higher survival rate during the first few months of life (Calambokidis and Gentry 1985; Chapter 5). Of course fetal growth is probably not independent of the mother. If the cost of producing a successful daughter is less than the resources the mother has she may withhold some to improve her chances of reproducing the following year. 2.5.4  Conclusion  It can be concluded that the size of the fetus is influenced by the size, age, and parity of the mother and that male fetuses grow faster and larger than female fetuses. Mothers that were pregnant in previous years produce bigger offspring than females carrying their first pup. Also, older females tend to carry progressively bigger fetuses  Chapter 2  Fetal Growth  34  until the age of 10-11 y. As females continue to grow beyond age 11 y, there is a senescent decline in the mass and length of their pups at birth. These observations are unlikely restricted to northern fur seals alone, but can undoubtedly be applied to other species of fur seals and perhaps to other pinnipeds as well.  2.6  Summary  Sex specific growth curves are described for northern fur seal fetuses. The relationships between body length, body mass, and gestational age are derived by regression analysis based on 7000 fetuses collected during 1958-74 as part of a joint Canadian - American pelagic research effort. Male fetuses grow faster and larger than female fetuses. Length approaches an asymptote with time, but the increase in fetal mass appears exponential until parturition. The size of the fetus is influenced by the age, size, and reproductive history of the mother. Primiparous females produce smaller pups than multiparous females. This difference in fetal size is presumably due to physiological changes associated with having been previously pregnant and is not explained merely by differences in the size and age of the different parities. Older and larger females produce progressively larger fetuses until reaching their reproductive prime at about the age of 10-11 y. Adult females continue to grow beyond this age, but there is a senescent decline in the length and mass of the young they carry. There is no indication that the sex ratio differs from unity either between months, across years, or between mothers of different ages and parities.  Chapter 3  Changes in Fetal Growth and the Condition of Pregnant Females from 1958-72 3.1  Introduction  Fetal growth and development are the outcome of many factors. Studies of humans and domestic animals have shown that fetal growth is influenced by 'intrinsic' factors such as the size, age and parity of the mother (e.g. Walton and Hammond 1938; Dickinson et al. 1962; Miller and Merrit 1979; Price and White 1985). They have also shown that the size of the fetus is 'externally' affected by the quality and quantity of food available to the mother (e.g. Wallace 1948; Hight 1966; Payne and Wheeler 1967b; Thorne et al. 1976; Mellor 1983; Hoist et al. 1986). In this study the size of northern fur seals at birth is estimated from a regression analysis of over 2 600 fetuses collected from 1958-72. The goal is to see whether changes in fetal growth can be detected among years and related to the build up of commercial fishing that began in the early 1960s. A second goal is to determine the average 'condition' of pregnant females for each year from 1958 to 1972. Condition indices describe the relationship between body mass and length, and have been used to compare the effects of biotic and abiotic factors on the well-being  35  Chapter 3  Changes in Fetal Growth & Female Condition  36  of individuals or the population. Individuals with more mass at a given length are assumed to be in better condition. This technique has been extensively applied to fish populations (see Gershanovich et al. 1985; Bolger and Connolly 1989; Cone 1989), but apparently only once to pinnipeds (Boyd and McCann 1989), and rarely to other species.  3.2  Materials and Methods  Pregnant females were collected from 1958-74 by Canada and the United States as members of the North Pacific Fur Seal Commission (Lander 1980a). A description of the procedures used to measure and age the fetuses and adults, and the biases inherent in the data set are discussed in Chapter 2. Maternal age, size, and parity influence the size of a northern fur seal fetus (Chapter 2). To minimize the effects for interannual comparisons, annual growth curves and condition indices were calculated for multiparous females between the ages of 8 and 13 y. The data were further limited to the months of February through June to reduce the effect of incomplete mixing of the seals during their annual migrations. Although data were collected until 1974, there were insufficient samples taken in 1973 and 1974 to estimate female condition and the average fetal size (see Table 2.1). 3.2.1  Female C o n d i t i o n Index  The physical condition of a single pregnant female was quantified with a condition index (CI) defined as the ratio between the mother's mass (M ) and her expected m  mass (M ) adjusted for fetal growth such that m  m  CI =  (3.1)  m and  M = M' -aMf m  m  .  (3.2)  Chapter 3  Changes m Fetal Growth & Female Condition  37  where M' is the mass of the mother (including fetus), and Mf is the mass of the m  fetus. The coefficient a corrects for additional mass associated with growth of the placenta, uterus, increased blood volume, etc.. The expected mass of the mother, M , is estimated from the allometric relationship between length and mass, which is m  derived by regressing the mother's adjusted mass (Eq. 3.2) against her body length using all 2 623 multiparous females, aged 8-13 y, collected from February to June, 1958-72. Thus, expected mass is predicted from body length. If a is chosen correctly, the mean condition of all females (from all years) will remain stable (i.e. CI = 1.0) throughout the period of fetal growth. Different values of a were tried systematically. For a given a, a condition index was calculated for each of the 2 623 females and plotted against the date each female was sampled. If a linear regression fit to the data did not have an intercept of one and a slope of zero, the a was rejected and another value tested. An annual estimate of female condition is the mean of the condition indices for all individuals collected in any given year. Thus on an annual basis, the estimated mean index of condition will fall about the grand mean, depending upon whether conditions in any given year were 'good' (CI > 1.0) or 'bad' (CI < 1.0). The terms 'good' and 'bad' are relative, not absolute. For example the entire population could experience poor conditions from 1958-74, but some years may be worse than others. 3.2.2  A n n u a l Fetal G r o w t h  The relationships between length and gestational age, and between mass and gestational age were linearized. Linear models were fit to samples of male and female fetuses to determine the average growth rate by sex for each year from 1958-72. The gestational age of the fetus was assigned according to the date when collected at sea. The date of implantation is believed to occur in early to mid November, and is thought to vary little among years and individuals (Craig 1964). Thus the fetuses  Changes m Fetal Growth & Female Condition  Chapter 3  38  Table 3.1: Changes in the mean mass (in kg) of mothers and pups (males: n = 8, females: n = 11) weighed at term (from Costa and Gentry 1986). The mothers were weighed pre- and post-partum and pups were weighed at birth. The coefficient a was calculated by rearranging the terms of Eq. 3.2 such that a = (M'm — Mm)/Mj.  Sex of Pup Carried  Mother's Mass Pregnant Post-Par turn  M'  M  52.5 49.7  42.3 40.4  m  male female  m  Pup Mass  a  6.23 5.10  1.64 1.82  Mf  were assigned a gestation age of 0 on November 10 (i.e. t = 0 d). Growth curves were therefore constructed for the period t = 82 (February 1) to t = 231 (June 30). The size (mass and length) of pups at birth (i.e. on June 30th) were estimated from these growth equations. There appears to be very little annual variability in the return of adult females to the Pribilof Islands and in the timing of birth (Gentry and Francis 1981; Chapter 4). Pupping begins on about June 20 and finishes about 40 days latter (Bartholomew and Hoel 1953; Peterson 1965). Over 50% of the pups are born during the first two weeks of July (Chapter 4).  3.3  Results  3.3.1  Female C o n d i t i o n Index  Various a levels were tested before selecting a = 1.75. This value gave a mean condition index of 1.0 when the data from all years were combined (Fig. 3.1) and is very close to the a value estimated from changes in the mean mass of pregnant females and pups weighed at term (Table 3.1). The predicted mass of adult females (in kg and adjusted for the size of the fetus) was estimated from the mother's length, L  m  (in cm), according to the linear relation M' = —45.150 + 0.643 L m  m  .  There was a general improvement in the condition of females sampled from 1958  Chapter 3  Changes in Fetal Growth & Female Condition  39  2.01.5-  c o  1.0-  C  0.5-  o  0.0-  F  M  A  M  J  J  Date (d) Figure 3.1: Female condition versus the date when collected (Feb. 1 - Jun. 30) for a = 1.75. A small amount of random variation was added to each data point in the figure to reduce overlap associated with roundoff error. All females, collected from 195872, were multiparous and between the ages of 8-13 y. The data were smoothed with lowess (solid line) and show little change in the condition index as the fetus develops over time.  to 1964 (Fig. 3.2 left panel). In 1965, the condition index dropped below 1.0 and remained relatively constant until 1972 when another sharp drop occurred. Females carrying male fetuses were in poorer condition than those with female fetuses (right panel of Fig. 3.2). It simply costs a mother more to produce a male fetus than a female fetus. 3.3.2  A n n u a l Fetal G r o w t h  Various transformations were tried to linearize the fetal length/time and the mass/time relationships before selecting the linear models (Fig. 3.3).  L = a + bt and M^ ' t  5  2 75  = a + bfr  The reciprocal power (2.75) was drawn from the relationship between  fetal length and mass (Chapter 2). Sex specific growth curves are shown in Fig. 3.4. Estimates of the average fetal size for two of the years were rejected after examining the regressions. In one case, 1963, the few data represented about two days of sampling and gave a poor fit (Tables 3.2 and 3.3). In the other case, 1967, there were also very few fetuses sampled. Sex specific regressions for male and female  5  Changes in Fetal Growth & Female Condition  Chapter 3  i—i—i—i—i—i—i—TT 1958  1962  1966 Year  1970  0.90  1 0.95  1 1.00 Male CI  1  r  1.05  1.10  40  Figure 3.2: Mean condition of multiparous females, ages 8-13 y, carrying male and female fetuses (left panel). The vertical bars are 95% confidence limits and the horizontal reference line shows CI = 1.0. The right panel plots the condition index of mothers carrying female fetuses (Female CI) against the index for those carrying male fetuses (Male CI). The solid line is the linear regression fit to the data that are identified by year. The dashed line is Female CI = Male CI.  fetuses were compared with regressions for the sexes combined. In all years except 1967, estimates of the mass and length of fetuses at term from the pooled samples fell between those of the males and females estimated independently. But in 1967, the average sized fetuses exceeded the estimated mass of males and females derived separately from sex specific regressions, and puts the reliability of the 1967 data into doubt.  3.4 3.4.1  Discussion Female C o n d i t i o n  The proposed condition index is similar to the relative weight procedure of Wege and Anderson (1978) which relates observed mass to a hypothetical ideal mass based on length. Cone (1989) points out that the relative weight procedure depends upon the determination of the assumed ideal relationship and is doubtful that such a relationship exists for all populations of a species (i.e. it is unlikely that the slope  Changes in Fetal Growth & Female Condition  Chapter 3  2.5 -  BO  0.0 -  0  20 10  §  0.0-  41  33 CD  2. Q.  -1.0 9  11  13  15  9  Time (d )  11  13  15  Time (d )  5  6  Figure 3.3: Linear regression of fetal length and mass over time. The top panels show jitter plots of length (in cm) and mass (in kg- ) versus the Julian date when collected (with units t ). A least-squares regression line is superimposed on each data set. No systematic structure appears in the residual plots which were smoothed with lowess (bottom panels). The data represent 2 869 multiparous females aged 8-13 y that were sampled over the period February 1 - June 30 from 1958-72. 36  5  and intercept remain constant). He does suggest that this approach may be suitable if comparisons are made within size classes, although even this is not entirely sure. I have compared the condition of multiparous females (age 8-13 y) sampled between February and June. As such, it can be argued that the comparisons are restricted to a given size class. Furthermore, by pooling all of the data to determine the average relationship between length and mass, each hypothetical mass estimated for any given individual represents a deviation from the total.  Thus I feel the  condition index is valid because it reflects the condition of an individual relative to the condition of all others sampled. The way that I have defined the calculation of CI results in estimates falling about the grand mean of 1.0 (Fig. 3.2). K CI > 1.0 then females were heavier than  Chapter 3  Changes in Fetal Growth & Female Condition  "i—i—r—i—r—i—i—i—i—i N D J F M A M J J A S  Date (d)  ' i 0  T 20  1 40  r 60  Length (cm)  42  T 80  Figure 3.4: Growth curves for male (solid line) and female (dashed line) fetuses taken from multiparous females aged 8-13 y. The growth curves for each sex were derived from least-squares regressions of the data shown in Fig. 3.3.  expected (based on their length) and are presumed to have experienced relatively favorable conditions.  A CI  < 1.0 is interpreted to mean that conditions were  poorer than during other years of sampling. On an annual basis, the average index of condition calculated for each year, will fall about the grand mean (CI depending upon whether conditions in any given year were 'good' (CI  =  1.0),  > 1.0) or  'bad' (CI < 1.0). Thus it appears there was a general improvement in the condition of females sampled from 1958 to 1964 followed by a drop below 1.0 in 1965, which remained relatively constant until 1972 when another sharp drop occurred (Fig. 3.2). If the condition index reflects food availability it might be interpreted to mean that fish resources were more abundant or available to the pregnant females through the late 50s and early 60s, but were reduced during the late 60s and early 70s. It is not clear how sensitive the condition index is to changes in food availability. Female northern fur seals have very little body fat relative to phocids. It is therefore  Changes in Fetal Growth & Female Condition  Chapter 3  43  Table 3.2: Regressions of fetal length in cm as a function of time (i) in days such that L = a + bt and 82 < t < 231 (Feb. 1 - June 30). The fetuses were from multiparous mothers aged 8-13 y. All regressions (except 1963) were highly significant 5  t  (p < .0001). Year  Intercept  Slope  a  b  1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1958-72  -38.144 -41.111 -28.181 -38.092 -33.244 27.773 -42.486 -32.479 -37.708 -45.142 -34.041 -30.227 -37.932 -43.775 -40.100 -38.203  6.739 6.931 5.910 6.664 6.380 2.239 7.101 6.332 6.630 7.394 6.483 6.054 6.761 7.244 7.019 6.717  I<231  64.29 64.22 61.64 63.20 63.72 61.80 65.44 63.76 63.05 67.23 64.49 61.79 64.82 66.33 66.59 63.88  Sample Size 564 452 577 319 189 67 74 68 93 18 157 101 62 68 57 2869  2  F  0.925 0.909 0.811 0.886 0.894 0.045 0.592 0.690 0.856 0.971 0.955 0.888 0.921 0.858 0.862 0.934  6952.1 4496.5 2463.6 2461.2 1569.3 3.0 104.6 146.6 540.6 533.9 3314.2 784.3 700.4 399.7 343.3 40711.1  r  likely that the condition index applied to fur seals is not very sensitive to short term changes in food, but instead may be an integration of the growth conditions experienced over several years. 3.4.2  F o o d Availability and Fetal G r o w t h  Many studies of domestic animals have shown that the size of the fetus is affected by the quality and quantity of food available to the mother (e.g. Wallace 1948; Hight 1966; Payne and Wheeler 1967b; Thome et al. 1976; Mellor 1983; Hoist et al. 1986). In particular, the fetal growth rate is extremely sensitive to underfeeding rather than overfeeding because there is a physiological upper limit on growth. The effect of food restrictions is felt more during the later part of pregnancy than during early  Changes in Fetal Growth & Female Condition  Chapter 3  Table 3.3: Regressions of fetal mass (W ) t  that W  t  1 / 2  '  7 5  = a + bt  5  44  in kg as a function of time (t) in days such  and 82 < t < 231 (Feb. 1 - June 30). Thefetuseswere from  multiparous mothers aged 8-13 y. All regressions (except 1963) were highly significant  (p < .0001). Year  Intercept  Slope  a  b  1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1958-72  -1.4698 -1.2897 -1.1339 -1.2123 -1.1784 -0.2183 -1.6765 -1.1364 -1.2050 -1.4624 -1.1033 -0.9606 -0.9984 -1.3125 -1.3379 -1.2713  0.228 0.212 0.200 0.206 0.205 0.139 0.240 0.200 0.205 0.232 0.199 0.186 0.188 0.213 0.216 0.211  W 3i 2  6.754 6.069 5.963 6.060 6.120 5.829 6.518 5.914 5.896 7.284 5.985 5.540 5.520 6.124 6.291 6.190  Sample Size 563 452 578 320 182 69 73 68 93 18 157 101 62 68 57 2864  r  2  0.949 0.918 0.884 0.902 0.927 0.216 0.741 0.806 0.895 0.964 0.965 0.915 0.928 0.846 0.909 0.952  F 10501.5 5065.2 4387.3 2921.3 2283.0 18.4 203.1 273.5 777.7 428.0 4306.2 1064.4 767.9 361.7 550.1 56315.8  development. During early pregnancy, females can draw on their own body reserves to met the requirements of their growing fetus. But during the last trimester of pregnancy, the females must rely on an external supply of good quality food to nourish their fetuses (Mellor and Murray 1982; Mellor 1983; Hoist et al. 1986). Undernourishment has been shown to cause smaller reductions in the length of fetuses than in their weight (Mellor 1983). This likely holds true for northern fur seals, given that increases in length and weight occur primarily during early and late pregnancy, respectively (see Chapter 2). There was considerable variability in the estimated average size of fetuses at term (Fig. 3.5). The general trend from 1958-72 was a relatively stable fetal length  Chapter 3  Changes in Fetal Growth & Female Condition  45  (weighted linear regression F\^\ = 2.13, p = 0.17), but a decrease in mass (-Fi,ii = 5.68, p = 0.03). If this trend is indeed real and fetal growth reflects feeding conditions, then inferences might be drawn about the food conditions experienced by the mothers in the North Pacific. Length is developmental and increases primarily while the mother is off the coasts of British Columbia and the USA. Thus it could be inferred from the estimates of length that females experienced satisfactory feeding conditions from 1958-72. Mass on the other hand is put on primarily as the seals migrate through the Gulf of Alaska and the Bering Sea on their way to give birth on the Pribilofs. The declining trend in mass at birth suggests reduced food availability through the 1960s. In general the estimated length and mass of fur seal fetuses at birth compare favorably to those predicted from the length - mass relationship previously calculated (Fig. 3.5 right panel). The exceptions are 1958 and 1960 when fetuses were much heavier than expected and 1970, 1971 and 1972 when fetuses were considerably lighter than expected. These 'good' and 'bad' years might reflect general feeding conditions in the Gulf of Alaska and Bering Sea. The low body weight relative to fetal length in the last 3 years of the study (1970-72) is particularly intriguing (Fig. 3.5 right panel). 3.4.3  Changes in the F u r Seal F o o d Base  Northern fur seals feed primarily upon herring  (Clupea pallasi), capelin (Mallotus  villosus), sandlance (Ammodytes hexapterus), sablefish (Anoplopomafimbria),walleye pollock (Theragra  chalcogramma) and squid in the Northeast Pacific (Perez and  Bigg 1981, 1986; Kajimura 1985). As the seals migrate northward through the Gulf of Alaska and into the Bering Sea their diet switches from primarily capelin, herring and sablefish to pollock and squid. They appear to feed opportunistically upon the most abundant schools of small fish.  Chapter 3  Changes in Fetal Growth & Female Condition  Year  Length (cm)  Figure 3.5: Estimated size of the fetus at term (June 30) from 1958-72. The estimates were derived from the linear regressions contained in Tables 3.2 and 3.3. No estimates are shown for 1963 and 1967. The right-hand panel plots the estimates of mass at birth against the estimates of length. The points are identified by year. Superimposed is the predicted relationship between fetal mass and length (taken from Chapter 2).  Many of the fur seal's prey species (particularly pollock and herring) are targeted by commercial fisheries as discussed in Chapter 16. Commercial ground and pelagic fisheries began in the Gulf of Alaska and Bering Sea in the early 1950s and peaked in the late 1960s and early 1970s. It is not clear what impact fisheries may have had on the fur seal food base (Chapter 16). But commercial fisheries are not alone in affecting the size of fish stocks. Major 'natural' fluctuations in abundance have been noted for a number of different pelagic species (see Chapters 9 and 11). These large changes in magnitude appear to occur over many years and may even have a regular periodicity of 8-12 years. Some have suggested the fluctuations might be  46  Chapter 3  Changes in Fetal Growth & Female Condition  47  explained by changes in sea surface temperatures and solar activity (Laevastu 1983; Bulatov 1989). Overall, little is known about the size of fish stocks during the time that fur seal fetuses were sampled. It is therefore not possible to relate fish abundance to the reported changes in fetal growth. Nevertheless it can be concluded that fish stocks do vary in abundance as a result of commercial fisheries and natural environmental fluctuations. These factors acting either alone or together have the potential to alter the quantity and quality of prey available to pregnant fur seals and hence could affect fetal growth.  3.5  Conclusion  Fetal growth, particularly mass, is likely a very sensitive indicator of feeding conditions during the last trimester of pregnancy while the fur seals pass through the Gulf of Alaska and Bering Sea on their way to give birth on the Pribilof Islands. This is unlike the female condition index which is probably an integrator of conditions experienced in previous year while feeding over the entire North Pacific migratory range. Thus it would seem that the overall feeding conditions for pregnant females in the North Pacific were much better in the early 1960s than in the latter part of the decade.  3.6  Summary  Annual estimates of the condition of pregnant northern fur seals and the size of their fetuses were determined from over 2 600 samples collected in the North Pacific during 1958-72. The condition index related the mass of the pregnant female to her hypothetical mass based on her length, correcting for the size of the developing fetus. The mean condition of pregnant females improved from 1958 to 1964, but dropped sharply in 1965 and remained low until 1972 when there was again another  Chapter 3  Changes in Fetal Growth & Female Condition  48  sharp drop. Fetal mass showed a general decline throughout the 1960s, but length appeared to be relatively constant.  A significant proportion of the final length  of a fetus is established early in pregnancy when the female can draw on her body reserves to meet the requirements of her growing fetus. But during the last trimester of pregnancy when mass increases most rapidly, the female depends on an external supply of good quality food to nourish her fetus. It is suggested that fetal growth, particularly mass, is likely a very sensitive indicator of feeding conditions during the last trimester of pregnancy while pregnant females are in the Gulf of Alaska and Bering Sea. This is unlike the female condition index which is likely affected by conditions experienced the previous year while feeding over the entire North Pacific migratory range. Thus it would seem that the overall feeding conditions for pregnant fur seals in the North Pacific were better in the early 1960s than in the latter part of the decade.  Chapter 4  Reproductive Synchrony and the Estimation of Mean Date of Birth from Daily Counts of Pups 4.1  Introduction  The mean date of birth is a useful parameter to estimate. It can be used to assign mean ages to pups or to compare the timing of birth between years and between different populations.  Other uses of mean date of birth include estimating pup  production when only a fraction of the pupping season is observed (B. Le Boeuf, pers. comm.) and testing hypotheses about interutero mortality and the timing of the previous year's breeding (Stirling 1971a). The best estimate of mean pupping date is the arithmetic mean of the numbers of pups born each day. Unfortunately following individual animals and recording births as they occur is difficult at the best of times and near impossible on high density breeding areas, so that obtaining a random sample to estimate this mean is not generally practical. A simpler approach to quantify reproductive timing is to choose a portion of the breeding area and count all the pups present. Mean pupping date for this portion can be estimated from cumulative daily counts of pups made 49  Chapter 4  Reproductive Synchrony & Mean Date of Birth  50  over the breeding season. In this chapter, I present and apply two methods for estimating the mean date of birth from daily counts of northern fur seal pups made in 1951, 1962, 1963 and 1983. I use the cumulative pup counts to describe and compare reproductive synchrony on the Pribilof Islands. My intent is to estimate the mean date of pupping for northern fur seals and to test if pupping patterns remained the same from year to year.  4.2  Methods  4.2.1  Sigmoid M e t h o d  The first procedure is essentially a curve fitting routine that assumes the number of pups P counted per day t on a study site, increases in a sigmoid fashion. I explored t  two sigmoid equations, the logistic  P = A/(l + ce~ ) Kt  t  (4.1)  and the Gompertz equation  (4.2)  P = Ae- ~ ' t  ce  K  where A, c, and K are the asymptote, the constant of integration, and the growth rate constant, respectively (Zach et al. 1984). These equations are used purely for data fitting, not as an indication of underlying population mechanisms. Note that the number of pups that die between counts is ignored at present and that immigration and emigration of pups are assumed to be negligible. The median date of birth (B) for each study site is estimated at one half the asymptote, such that P = A/2 where t = B. t  Making these substitutions into  Eqs. 4.1 and 4.2 yields  B =  (4.3)  where the constant e equals 1.0 when solving for the logistic equation and In 2 (= 0.69312) for the Gompertz equation. Fifty percent of the pups are born before day  Reproductive Synchrony & Mean Date of Birth  Chapter 4  51  B. In the case of the logistic equation which is symmetrical, B is both the mean and median birth date; but when applied to the Gompertz equation, B is just the median. The variance of B is estimated by the delta method: n  var(/[pi  ,p ,... p ]) = 2  n  where n is the number of parameters, with respect to parameter j>,, a  pi  n  2  df Of Qf. of. Pi PJ a  a  PpiPJ  (-) 4  J^- is the partial derivative of f(pi,P2->---Pn)  is the standard deviation of p,-, and p . p{p  coefficient of correlation between pi and pj. Thus, for Eq. 4.2, since ^ In c)  4  is the  = (In e —  K~ and ^ = {cK)~ it follows that 2  x  ,„\  var(5)=  ( 1 \ ^_j li 2  a +2  aa p +  „lne-lnc c  K  cK  j  /lne-lnc\ R2  2  , „. (4.5) f l  Estimates of A, c and K are obtained by fitting the logistic and Gompertz equations to the daily pup counts using a nonlinear curve fitting routine such as NONLIN in the Systat (1989) statistical package. The output of NONLIN includes standard errors and a correlation matrix of parameters needed to estimate the variance of B. 4.2.2  Summation Method  The second procedure for estimating the mean date of pupping is referred to as the 'summation' method. It assumes the probability, p(t), of a pup being born on day t at a given study site is =  (4-6)  where PQ = 0, and Pp is the maximum number of pups counted at the end of the period t = 1... T. The mean date of birth is derived from T  B = Y,tp(t) t=i  Reproductive Synchrony & Mean Date of Birth  Chapter 4  T  52  T  t=i (4.7) The last step in Eq. 4.7 is obtained by expanding the summations, combining like terms and using the fact that PQ = 0. Missing values of Pt are estimated from  Pt =  Pt-i + P i t+  2  For two or more consecutive missing counts, the number of pups can be approximated by linear interpolation. The variance of B for the summation procedure is also estimated by the delta method (Eq. 4.4). But instead of two parameters as in the sigmoid method, there are now T parameters to contend with. Partial derivatives of Eq. 4.7 are Jp^- —  EfcTi t Pf 1p  2  311(1  = -Pf for i = 1,... T -1. The daily counts, P , are assumed 1  to be binomially distributed with variances a\  t  = P (l t  - PtPj ). 1  Correlation  coefficients are probably low, but since they are not known they are conservatively set to 1.0. Variance of Pj is also not known and is assumed to equal the variance in the number of pups counted at the end of the breeding season when pup counts plateau. Estimating the variance of B in this manner should provide a reasonable indication of the reliability of B. If no pups die and there are no errors in counting, the numbers of pups counted each day will rise continuously to an asymptote.  Under these conditions Pj  (the  denominator in Eq. 4.7) is the maximum number of pups counted at the end of the period  t = 1,... T. Since counting errors do occur, the ratio Pt/Pr will exceed 1.0 if  the maximum number of pups counted occurs before time T. One way to correct for  Chapter 4  Reproductive Synchrony & Mean Date of Birth  53  this is to set Pr = max P for t = 1,... T -1. The variance of PT will approximately t  equal the sample variance in the numbers of pups counted over the last few days of the breeding season. A computational difficulty with the summation method is that values of p(t) can be negative because of errors in counting pups (see Eq. 4.6). This is not considered a major problem however because the final formulation of B (Eq. 4.7) is expressed in terms of the summation of daily counts. Such a derivation tends to reduce the effects of counting errors on the estimate of B. Another point to consider in both the sigmoid and summation methods is how to interpret B. Presumably counts are made at the same time each day during a breeding season. If for example counts are made at 2 pm each day then B = 1 would refer to 2 pm of day 1 and B = 1.5 would be 12 hours later (i.e. 2 am on day 2). Estimates of B are therefore based upon the time of day pups were consistently counted during the breeding season. 4.2.3  Pup Mortality  The above methods assume that no pups died between observations. This means both procedures calculate the mean birth date of pups that were alive on the days of counting, thereby resulting in an under-estimate of the true mean date. This can be corrected if dead pups (Dk) are counted on days k = 1... T. If dead pups are counted daily, the expected number of pups (P ) that would t  have been counted on day t if all had survived is (4.8)  k=\ which is the count of live pups, plus the daily accumulation of dead pups lying on the beach. If dead pups are not counted each day, but the mortality rate m is known, the total number of pups  (JDtotal)  wovld have died over the period t = 1... T can  Chapter 4  Reproductive Synchrony & Mean Date of Birth  54  be calculated from 771  Aotal = z  (4.9)  -Palive  1—  m  where .Palive is the number of live pups counted at the end of the breeding season, and m is the fraction of pups born that were dead on day T . The number of pups (Dk) that die on day k will equal the product of the total number of dead pups (-Dtotai)  and the daily mortality rate. Pup mortality curves can be constructed  from periodic counts of dead pups (see Results for example). If the daily mortality pattern is not known, the number of pups that died each day can still be estimated by prorating the total number that died from t — 1...T according to  Dk = ^  Pk T  „  Aotal  •  (4.10)  Substituting Eqs. 4.9 and 4.10 into Eq. 4.8 yields  Pt=Pt+  Palive  m  Ej=i Pj  •z——— tr ' 1 - m E L i Pk  •  C 4  1 1  In this formulation, the expected pup count is sensitive to errors in .Palive the 5  number of pups alive at the end of the breeding season. One means of improving the estimate of P a l i v e is to let it equal the mean number of pups counted over the last few days of observations. 4.2.4  F u r Seal P u p Counts  Northern fur seal pups were counted on study sites from June to August in 1951 (Kitovi amphitheatre - Bartholomew and Hoel 1953), 1962 and 1963 (Kitovi study site - Peterson 1965), and 1983 (East Reef study site - unpublished data from the files of the National Marine Mammal Laboratory, Seattle, WA). No other daily counts of northern fur seal pups are available. The Kitovi sites are located on St. Paul Island and the East Reef site is on St. George Island. In 1983 the pups were  )  Chapter 4  Reproductive Synchrony & Mean Date of Birth  counted daily between 11 am and 12 noon from an observation blind overlooking the rookery (R. Gentry, pers. comm.). Unfortunately written records do not indicate what time of day pups were counted at the other study sites, although I think it can be safely assumed that pups were consistently counted at the same time each day within a given year.  In all likelihood counts were made at about mid-day.  Nevertheless counts between years might vary by as much as 8 hours depending upon whether they were made at the beginning of the day or at the end of it. The daily mortality pattern is estimated from counts of dead pups made by Kenyon et al. (1954) in 1951 on Vostochni rookery. The study site was bounded by painted rocks and covered an area of 15 000 square feet. Accumulated carcasses were counted from an elevated walkway that bisected the study site. Counts began when the first pup was born and continued at 5-day intervals.  4.3  Results  The number of pups counted on the four study sites are superimposed with Gompertz curves in Fig. 4.1. The numbers of pups born each day was derived from the Gompertz model and was scaled by the maximum number born and expressed as a relative number in Fig. 4.2 to facilitate comparing the four data sets. Model parameters and the estimated mean dates of birth are reported in Tables 4.1 and 4.2 for the sigmoid (Eq. 4.3) and summation (Eq. 4.7) methods, respectively. Fitting sigmoid curves to the four sets of 'uncorrected' pup counts suggests that the mean date of birth lies between July 3-5 and July 9. The earlier dates are based upon the 1962 and 1963 counts done by Peterson (1965); while the later date is estimated from the 1951 survey of Bartholomew and Hoel (1953) and the 1983 NMML data set. The logistic and Gompertz models gave good fits to the daily pattern of pup counts, and indicate a median date of birth on July 6. Because the results of both models are similar and the squared deviations of samples values were  55  Reproductive Synchrony & Mean Date of Birth  Chapter 4  Table 4.1:  56  'Sigmoid' estimates of the median birth date (i?) with 95% confidence  intervals for pups born on 4 study sites in 1951, 1963, 1964, and 1983 where the day B = 1 represents June 15. Pups were counted on n days ending day T.  The estimates  were derived using the raw daily counts (uncorrected) and the daily counts corrected for annual pup mortality rra. Model parameters A, c and k, and associated standard errors are from the Gompertz model (Eq. 4.2) fit using NONLIN (Systat 1989). The correlation between c and k is p k and MS is the mean squares of the residuals. The c  average number of pups counted in the last five days of counts is i a l i v e difference between uncorrected B and mortality corrected B equals  Sigmoid Model Uncorrected  Parameter A (se) c (se) Jb (se) MS Pck  B (<r ) 95% CI B date A (se) c (se) Jb (se) MS B  Corrected  Pck  •'alive m B (<x ) 95% c i B date n T AB B  Both  1951 Kitovi Amp. 799.42 (6.03) 18.56 (1.36) 0.133 (0.022) 212.47 0.975 24.70 (1.01) 22.74 - 26.68 July 9 925.43 (8.35) 13.70 (0.86) 0.133 (0.003) 241.12 0.973 768.24 0.158 26.37 (1.27) 23.88-28.85 July 10 40 52 1.67  1963 Kitovi 707.91 (10.30) 29.43 (8.14) 0.193 (0.015) 1152.66 0.984 19.44 (1.81) 15.89 - 22.98 July 3 784.5 (12.22) 18.77 (3.71) 0.162 (0.011) 1174.61 0.977 683.40 0.124 20.43 (2.06) 16.39 - 24.46 July 4 41 52 0.99  1964 Kitovi 739.24 (11.84) 22.90 (4.67) 0.169 (0.011) 1011.94 0.976 20.66 (1.85) 17.03 - 24.28 July 5 804.47 (13.47) 17.48 (3.07) 0.150 (0.009) 929.60 0.977 724.80 0.076 21.50 (1.96) 17.65 - 25.35 July 5 37 52 0.84  >d the  ar  AB.  1983 East Reef 319.95 (14.74) 9.63 (1.85) 0.104 (0.011) 266.92 0.967 25.37 (6.06) 13.50-37.24 July 9 328.59 (16.05) 9.39 (1.81) 0.101 (0.011) 265.39 0.969 286.82 0.036 25.70 (6.44) 13.08 - 38.31 July 10 42 45 0.33  Chapter 4  Reproductive Synchrony & Mean Date of Birth  • I ' I ' I ' I ' I—i——' I ' I ' I ' I '  I—  1  20  30  June  10  20  July  30  20  30  June  10  20  July  57  r  30  Figure 4.1: Daily number of live pups counted on 4 study sites from June 15 to August 5. Data are from Kitovi amphitheatre (1951), Kitovi Study Site (1962 and 1963), and East Reef Study Site (1983). The field counts were fit with Gompertz curves (Eq. 4.2, Table 4.1). An estimated median date of birth (July 6, t = 22.53) is shown by the dotted vertical line.  slightly smaller for the Gompertz model, Table 4.1 only contains Gompertz results. The summation method suggests the mean date of birth occurs 2 days later on July 8. The later date reflects the sensitivity of the summation estimates to the magnitude of Pj in Eq. 4.7. Unlike the sigmoid curves which are less sensitive to outliers, the summation method is sensitive to errors in the maximum number of pups counted by an observer. The reliability of the estimate B resulting from the sigmoid procedure or the summation method depends upon the accuracy of the daily counts during the later part of the season when peak numbers of pups are present. Reliability can be affected by prevailing weather conditions, such as wind  Reproductive Synchrony & Mean Date of Birth  Chapter 4  58  Table 4.2: 'Summation' estimates (Eq. 4.7) of the mean birth date (J?) with 9 5 % confidence limits for pups born on 4 study sites in 1951, 1963, 1964, and 1983 where day B = 1 represents June 15. The estimates were derived using the raw daily counts (uncorrected) and daily counts corrected for the annual mortality of pups on land, m. The maximum number of pups counted from day 1 to day T (the last day pups were counted) is PT- The standard deviation of Pj is set equal to the standard deviation of n days of counts made from July 25 to August 5; and the average number of pups counted in the last five days of counting is Paiive- The difference between uncorrected B and mortality corrected B equals AB. Summation Model Uncorrected  Parameter PT{<TP ) B (<r ) 95% c i B date Pr{<rp ) 'alive m B (<T ) 95% CI B date n T AB T  B  Corrected  T  B  Both  1951 Kitovi Amp. 781.00 (8.00) 25.10 (0.16) 24.79 - 25.40 July 9 890.01 (15.12) 768.24 0.158 26.55 (0.01) 26.53-26.57 July 11 12 52 1.45  1963 Kitovi 757.00 (44.53) 21.99 (1.23) 19.57-24.40 July 6 828.92 (47.11) 683.40 0.124 22.77 (1.15) 20.51-25.03 July 7 8 52 0.78  1964 Kitovi 777.00 (38.39) 22.65 (0.92) 20.84 - 24.46 July 7 827.10 (30.00) 724.80 0.076 23.02 (0.55) 21.94 - 24.09 July 7 6 52 0.37  1983 East Reef 341.00 (38.16) 27.18 ( 1.16) 24.90-29.46 July 11 346.64 (37.87) 286.82 0.036 27.27 (1.11) 25.09 - 29.44 July 11 5 45 0.09  and driving rains that cause pups to seek shelter behind logs and under rocks (Ohata and Miller 1977).  Revised estimates of the mean date of birth correcting for natural mortality are found in Tables 4.1 and 4.2. Annual estimates of mortality rates were taken from Trites  (1989).  The total number of pups that died over the period  t=  1...  T was  estimated using Eq. 4.9 and multiplied by the cumulative mortality curve in Fig. 4.3 to determine the numbers of dead pups counted on any given day. The results show that the birth dates estimated with the 'uncorrected' pup counts are underestimated by as much as 1.7 days and as little as 0.1 days depending upon the amount of natural mortality (the greater the mortality, the greater the bias).  The median date of birth over the 4 years of study was July 7 based on  Chapter 4  Reproductive Synchrony & Mean Date of Birth  '  I • I • I ' I • I 20  30  June  10  20  July  30  1  i 20  1  i 30  June  1  i 10  1  i 20  1  July  i— 30  59  r  Figure 4.2: Relative number of pups born each day on 4 study sites from June 15 to August 5. Data are from Kitovi amphitheatre (1951), Kitovi study site (1962 and 1963), and East Reef study site (1983). The number of pups born each day was derived from the fitted Gompertz model (Table 4.1, Fig. 4.1) and scaled for the purpose of comparison. An estimated median date of birth (July 6, t = 22.53) is shown by the dotted vertical line.  the sigmoid methods corrected for mortality. The 'corrected' summation method suggests a mean date of July 9. Pup counts were also corrected by the prorating method (Eq. 4.11) and compared with the mortality curve method. Results of the two mortality correcting procedures differed very little. Differences in the mean date of birth for both sets of corrected pup counts only varied between 0.01 and 0.10 days. Summation estimates of B have much tighter confidence intervals than the sigmoid estimates. However, the summation method only includes binomial variability and does not take into account daily miscounting errors. Both methods are sensi-  Chapter 4  Reproductive Synchrony & Mean Date of Birth  500  60  Number of Dead Pups  -  0  30 June  10  20 July  30  o o i  10 August  Figure 4.3: Cumulative total of dead pups counted on a portion of Vostochni rookery in 1951. The solid line is a logistic equation: dead pups = 455.61(1 +  88.57 -o.i53(day-i3))-i e  w n e r e  d  a  y  _j ^j  u n e  1  5  a n c  | dead pups = 0 for day < 12.  The dashed curve shows the number of pups that died each day. An estimate of the mean date of birth in 1951 (July 11, t = 26.53) is shown by the dotted vertical line.  tive to counting errors at peak densities. For the sigmoid method, counts must be continued long enough in the breeding season to establish the asymptote. Counts in 1983 stopped five days earlier than in other years which increased the error associated with fitting the sigmoid curve to the data, thereby increasing the variance of the estimate of B. In the case of the summation method, which depends upon the maximum number of pups counted, care must be taken to count the pups as accurately as possible at peak numbers. The general pattern of births were compared among years by contingency analysis (using only days when pups were counted in all four years) and found to be significantly different (Pearson X57 = 425.3, p < 0.001). Kolmogorov Smirnov tests suggest the difference between 1951, 1963 and 1964 was not significant (ks= 0.15, n = 4, p = 0.95), but that the pattern of pupping in 1983 differed significantly from previous years (ks= 0.65, n = 20, p < 0.001). Mean and median birth dates were  Chapter 4  Reproductive Synchrony & Mean Date of Birth  deemed significantly different if their estimated confidence limits did not overlap (see Tables 4.1 and 4.2). But the only differences detected were among the summation estimates (confidence intervals on the sigmoid estimates were too wide). For the summation estimates, the difference between mean pupping dates in 1963 and 1964, and between 1951 and 1983 were not significant. However pupping in 1951 and 1983 did occur significantly later than in 1963-64.  4.4  Discussion  Analysis of harvest statistics and behavioural studies have provided information on the arrival times of adult and juvenile northern fur seals on the Pribilof Island rookeries (Jordan and Clark 1898; Bartholomew and Hoel 1953; Peterson 1965, 1968; Gentry 1981; Bigg 1986). Analysis of the cumulative pup counts adds further information about pupping and timing of the fur seal's life cycle. The number of pregnant females returning each summer to give birth on the small Bering Sea islands has ranged from 450 000 in the 1950s, to just under 200 000 in the mid 1980s (Lander 1980a; Roppel 1984; York and Kozloff 1987; Trites 1989). Reproduction has remained highly synchronized and consistent from one year to the next (Figs. 4.1 and 4.2). This phenomenon may reflect climatic seasonality and is likely a strategy that maximizes reproductive success (Peterson 1965, 1968; Lns 1990; Chapter 6). Parturition in northern fur seals occurs on average 0.8 days after arriving ashore and appears to be triggered by their arrival on the rookery (Peterson 1965; Bigg 1984; Gentry and Holt 1986). Thus the abrupt rise in the number of pups born per day in late June reflects the return of pregnant females (Fig. 4.2). The return schedule is positively skewed, indicating a tendency to be late rather than early. The late arriving pregnant females are probably young and primiparous with a less developed homing instinct than older females.  61  Chapter 4  Reproductive Synchrony & Mean Date of Birth  62  Pupping occurs over a 5 week period. Over 50% of the Pribilof pups were born during the first two weeks of July (56-77%) with over 75% born within a three week period (75-91% between June 28 - July 20). The general pattern of births was consistent in all 4 years studied. Further evidence that this pattern remains the same across years is that daily counts of females at East Reef Rookery peak during the week ending July 13 and have done so for the past 16 years (Gentry and Francis 1981; R. Gentry, pers. comm.). In 1951 pup mortality peaked on July 25 (Fig. 4.3), 15 days after the median birth date of July 11 (Kitovi 1951, Fig. 4.1, Table 4.2). Although the live and dead pups were not counted on the same study site, they were made in the same year and should be comparable. Female fur seals typically spend the first 7 days ashore nursing their pups before departing to feed at sea (Gentry and Holt 1986). During the mother's absence the pup receives no care from any other animal. In 1951, the first feeding trip to sea lasted an average of 7 days (a = 1.1, n = 11, Bartholomew and Hoel 1953; Gentry and Holt 1986). Comparing Figs. 4.1 and 4.3, it would seem that most pup mortality occurs while the pup is left unattended for the first time, with the greatest number of deaths occurring just prior to the female returning to nurse her pup. The small differences of 1-3 days in the mean dates of birth among rookeries were deemed significant from 1951 to 1963-64 to 1983. There are several reasons to expect differences in the timing of birth. For example slight differences in timing could exist between rookeries given that adult females show a high site tenacity to their rookery of birth, and the timing of birth may be genetically determined. Another consideration is that the age composition of the rookeries may have changed with time due in part to the commercial harvesting of females that occurred from 1956-68 (Lander 1980a). A rookery with many older pregnant females should show an earlier mean date of birth than a rookery of young females, because the older  Chapter 4  Reproductive Synchrony & Mean Date of Birth  63  females return first to the breeding islands (Bigg 1986). Thus the female harvest may account for the apparent shift in the numbers of pups born each day at Kitovi (Fig. 4.2), resulting in the mean date of birth shifting from July 11 (1951) to July 7 (1963-64).  4.5  Summary  Two methods for estimating the mean date of birth from daily counts of northern fur seal pups are presented and applied to data collected on the Pribilof Islands in 1951, 1962, 1963 and 1983. The mean date of birth over the 4 years was July 9. Reproduction is highly synchronized and consistent from one year to the next. Pupping occurs over a 5 week period with over 50% of the pups being born during the first two weeks of July.  Chapter 5  Thermal Budgets and Climate Spaces: The Impact of Weather on Pup Survival 5.1  Introduction  Historically, northern fur seals were believed to come ashore only to reproduce on the Pribilofs and a few other islands in the Bering and Okhotsk Seas (see Fig. 1.1). It was commonly thought they were restricted to the North Pacific by a thermoregulatory need for cool wet summers (Bartholomew and Wilke 1956). However, in 1968 a growing colony of northern fur seals was discovered breeding on San Miguel Island, California, under hot and dry conditions (Peterson et al. 1968b), which indicates that fur seals can tolerate a wider range of environmental conditions than was initially believed. Fur seals, like all pinnipeds, must cope with diverse climatic conditions during the terrestrial phase of their life cycle (Whittow 1987). Newborn fur seals on the northern and southern breeding islands appear to experience extreme thermal stress. In the Bering and Okhotsk Seas, rain in combination with wind and low air temperatures appear to bring pups to their lower limit of tolerance (Keyes 1965; Ohata and Miller 1976; Irving et al. 1962; Blix et al. 1979a) causing high pup mortality  64  Chapter 5  Thermal Budgets & Climate Spaces  65  (Roppel et al. 1963; Vladimirov 1974). In the south, on San Miguel Island, high air temperatures and low wind speeds axe linked with high mortality rates (Antonelis and DeLong 1985a,b, 1986). In the present study, I employ Porter and Gates's (1969) 'climate space' approach to investigate the impact of weather on the thermal balance of neonatal northern fur seals breeding on the Pribilof Islands and San Miguel Island. I also consider the ability of pups of a related species, the Galapagos fur seal  (Arctocephalus galapa-  goensis), to cope with the equatorial climate of the Galapagos archipelago. Climate space analysis is one of the simpler, well tested, biophysical models that offers insight into biophysical constraints on animals (Spotila et al. 1989) and has been applied to many species (e.g. Porter and Gates 1969; Spotila et al. 1972; Morhardt and Gates 1974; Morhardt 1975; Gates 1975, 1980; Campbell 1977; Scott et al. 1982; Christian et al. 1983; Spotila et al. 1990). I begin by outlining the general biology of fur seal pups and the thermal budget equation that accounts for the exchange of energy between an animal and its environment. After discussing model assumptions and choice of parameters, I plot a 'climate space diagram' containing all combinations of air temperature and solar radiation that a pup can withstand under various levels of humidity and wind speed. A sensitivity analysis shows the reliability of model predictions and the effect of faulty parameters. Finally, I consider the effect of weather on pup mortality, and the roles of behaviour modification and body size in light of the climate space diagrams.  Chapter 5  5.2  Thermal Budgets & Climate Spaces  66  Biology of the Pup  The northern fur seal is the second largest of 9 fur seal species (King 1983; Croxall and Gentry 1987), with a mean birth mass of 5.67 kg for males and 5.13 kg for females (Scheffer and Wilke 1953; Lander 1979a; Fowler 1990a). On the Pribilofs (St. Paul and St. George Islands), pups are born on 20 rookeries (Lander 1980a) consisting of sand or dark basalt rocks. Most pupping occurs from the last week of June until the third week of July with 80% being born during the first two weeks of July (Peterson 1968). Over the next three months, pups are fed a lipid-rich milk with a low water content for 2 days while their mother is ashore, and fast for the next 4-8 days when she is absent (Bartholomew and Hoel 1953; Costa and Gentry 1986). Most pup mortality occurs within the first few weeks of life (Kenyon et al. 1954) and is highly variable between years (Chapman 1961; Lander 1975, 1979b; Trites 1989). The San Miguel colony occurs 10 m from the sea on a wide sandy beach at Adams Cove. The presence of adult fur seals bearing tags from Soviet and American rookeries in the north indicates that the new colonies are the result of immigration from those areas (Peterson et al. 1968b). San Miguel pups are born during June and early July, with the mean date falling in the last week of June (Antonelis and DeLong 1985a). Their mass at birth is identical to those born on the Pribilof Islands (DeLong et al. 1981) with the majority of pup deaths occurring within 1 to 4 days postpartum (DeLong et al. 1982). The Galapagos fur seal is the smallest fur seal species, occurring only on the Galapagos archipelago (King 1983; Bonner 1984). Their mean mass at birth is in the order of 3.0 - 3.8 kg (Trillmich and Limberger 1985; Limberger et al. 1986; Trillmich 1987). Pups are born along a 20 m wide zone that runs parallel to the water, consisting of lava flows, caves and large boulders. Pupping begins in mid  Chapter 5  Thermal Budgets & Climate Spaces  67  August and lasts until mid November, peaking about the last week of September. The young are nursed until 2 years of age. Fur seal pups are insulated by fat, skin and fur. Heat loss is under cardiovascular control (Blix et al. 1979a) and is affected by the boundary layer of air next to the animal's surface. The fat layer at birth is thin (2-4 mm) in the northern fur seal and does not increase much over the first two weeks of life (Blix et al. 1979a). Subcutaneous tissues (skin and fat) provide little insulation (Irving et al. 1962). Unlike other pinniped pups, such as harp seals (Davydov and Makarova 1964; Blix et al. 1979b) and Weddell seals (Eisner et al. 1977), fur seals increase fur density rather than blubber thickness. Northern fur seal pups have a short dense fur with a mean length of underfur of 7.7 mm and guard hairs 16.3 mm (Scheffer 1962). The guard hairs make the fur stand up, which increases its thickness and insulative quality, and help to maintain the layer of air trapped in the underfur when the seal is immersed in water (Romanenko and Sokolov 1988). The insulation provided by the air is easily destroyed by wind and rain and is maintained by fluffing the flattened fur (Webb and King 1984). Adult-type underfur and guard hairs do not appear among the northern fur seal pup hairs until mid-August, prior to the September molt. At this time the color of the pup coat begins to change from black to silver. Under thermal neutral conditions, excess metabolic heat is easily dissipated across the body surface. However, when stressed by heat, fur seals will lower their metabolic rate, increase the flow of blood to their limbs and body surface, and begin panting (Bartholomew and Wilke 1956; Ohata and Miller 1977). At the other extreme, should pups begin to cool below their lower critical temperature, they will increase their metabolic rate and reduce the flow of blood to the appendages.  Thermal Budgets & Climate Spaces  Chapter 5  5.3  68  Methods  5.3.1  T h e r m a l Budget  The balance between heat input and heat loss for any endotherm maintaining a constant body core temperature is described by the thermal budget equation  ^  - e<r(273.15 + T f + M - XE h  p C p ( T b  The notation follows Campbell (1977), where  ~  T  a  )  +  m  b  (  M  "  A  ^ =0  (5.1)  Rab is the flux of short- and long-wave s  radiation absorbed by the animal from the environment per unit body surface area (Wm~ ). The second term «7(273.15+2b) is the long-wave emittance of the animal 2  4  (assuming T = Tb) where e is the surface emissivity, a is the Stephan-Boltzmann s  constant (5.67 X 1 0  Wm  -8  _ 2 o  K  - 4  ), T  B  is the animal's surface temperature and  Tb is the deep body core temperature (°C). The flux of metabolic heat to the body surface per unit area is M (W m ) . The flux density of latent heat lost through the -2  evaporation of water from the respiratory tract and skin of the animal is denoted by XE where A is the latent heat of vaporization and E is the water vapor flux  pCp(Tb-T )/rHa is the flux density of sensible heat from the  density. The fifth term  a  body assuming T = Tb, where p is the air density, Cp is the specific heat of air, T is s  a  the air temperature, and r// is the thermal laminar boundary layer resistance. The a  volumetric heat capacity of the air is a constant (pep = 1200 J m  - 3  ° K ) . The first - 1  five terms of Eq. 5.1 describe the heat exchange for an animal that has the same surface and deep body temperature and hence no internal resistance to heat loss. The sixth term where  rub{M-XE)/r is a correction factor for the assumption of T = Tb, e  s  rjjb is the resistance of the fur seal coat and peripheral tissue layer and r is e  the parallel equivalent resistance of rjja and the radiative transfer resistance (r ), r  where r  e  - 1  =  r/fa  - 1  +r  layer resistance in s m  r  _ 1  -1  and rr  = pCp/[4e<7(273.15-|-T6) ]. Note that the boundary  is estimated as rjja  3  = 307-^/d/u (Campbell 1977), where d  is the diameter (m) of the pup, and u is the wind speed (m s ) . _1  Chapter 5  Thermal Budgets & Climate Spaces  69  The thermal budget can be used to identify the combination of radiation levels and air temperatures that a fur seal pup can tolerate. The extreme conditions that a pup can withstand are determined by balancing Eq. 5.1 for an animal with a body temperature, metabolic rate, and evaporative water loss that are near upper and lower sustainable physiological limits. This assumes that there are thresholds above and below which animals die when so exposed. The lethal combination of radiation level and air temperature can be computed and plotted to bound the set of environmental conditions that are tolerable by a pup. Such a plot is referred to as a 'climate space diagram' (Gates 1980; Tracy 1982; Spotila et al. 1989). 5.3.2  Available Radiant Energy  The 'space' can be further bounded by excluding all combinations of environmental conditions that could never occur on the Pribilof Islands. These physical limits prescribe the minimum and maximum amounts of energy available at any air temperature and are derived as follows. The minimum radiant energy available to the fur seal pup is the nighttime longwave irradiance from the sky and the ground. Absorbed nighttime irradiance is calculated as a function of a, T , aj_, (the long-wave absorptivity of the seal), and e« a  (the average emissivity of the surroundings), such that  (5.2)  Night Rabs = a e a(273.15 + T ) L  e  a  4  The maximum flux of radiation that could be absorbed by a basking fur seal pup in direct sunlight at any temperature is equal to the direct, scattered and reflected sunlight at solar noon, plus the long-wave radiation. It is written as  Ap Sun Ra = a -^S + a Sd + Night Ra bs  e  p  s  hs  (5.3)  where a is the short-wave absorptivity, A is the surface area of the animal's body, e  Ap is the projected area of the animal's body on a plane perpendicular to the solar  Thermal Budgets & Climate Spaces  Chapter 5  70  beam, S is the direct beam solar irradiance, and Sd is the diffuse solar irradiance. p  5.4  Parameterization  The thermal budget equation contains many parameters and constants that need to be estimated. In many cases this means making simplifying assumptions about the uniformity of the physical environment, cloud cover, and substrate. It also means making sweeping generalizations about the size, body temperature, metabolic rate and physiology of a 'typical' newborn fur seal pup. The extreme combinations of radiation levels and air temperatures that a pup can tolerate are largely determined by the range of body temperatures, metabolic rates, and peripheral tissue resistance levels that a pup can sustain. In general there is a good understanding of the lower physiological limits of northern fur seal pups. Information on the upper limits is not as good; hence I must rely on data recorded for subadult northern fur seals, Galapagos fur seal pups, and other pinniped species. The following outlines how and why particular values were chosen to describe the physical and physiological characteristics of northern fur seal pups, and the environmental conditions that bound the climate space. 5.4.1  M a s s and Diameter of the P u p  Measurements made in the field indicate the average newborn northern fur seal weighs about 5.4 kg (Scheffer and Wilke 1953; Lander 1979a; Fowler 1990a). To my knowledge, the diameter of pups has never been recorded, but can be determined analytically by relating mass to surface area in the following manner. Based on a study of harbor seals, the relationship between surface area A (cm ) and mass W 2  (kg) is A = 800 W / (Irving et al. 1935). Recent calculations suggest this equation 2  3  can be applied to otariid seals without introducing major errors (Innes et al. 1990). Thus the surface area of a 5.4 kg fur seal pup should be 2 462 cm . Interestingly, 2  Thermal Budgets & Climate Spaces  Chapter 5  71  Blix et al. (1979a) found the surface area of an average size newborn, which they measured, was 2,447 cm , of which 40% was flippers. The diameter of the pup 2  (d) can be estimated by assuming the body core has a cylindrical geometry where d = 0.60 A (7rZ) . _1  Given the length (i) of a pup is about 62 cm (Scheffer and  Wilke 1953; Lander 1979a), then the average diameter should be about 7.6 cm. I compared this method of determining diameter with the measurements taken by Luecke et al. (1975) from two female seal lions, where flippers account for about 35% of total body area. The results suggest that the cylinder approximation could underestimate diameter by as much as 12.5%. Thus, failing any field measures, I believe the diameter of an average pup lies between 7.6 and 10.0 cm and use a value of 8.7 cm in subsequent calculations. 5.4.2  B o d y Core Temperatures  The minimum body core temperature that a pup can sustain is based on studies of pups exposed to cold air and immersed in water. In one study, the body temperature of a pup immersed in 10° C water dropped from 37.4° C to a stable 34.8° C after 20 minutes (Blix et al. 1979a). Pups immersed in ice water experienced a linear drop in body temperature for 30 minutes to 30°C, at which point they were removed. These studies suggest a minimum sustainable temperature of about 34° C which is probably accurate to within 1°C. While no studies have been done on heat-stressed pups, several body temperatures have been taken from hyperthermic subadult males that were immobilized and near death during the harvest roundup. In one study, Ohata and Miller (1977a) recorded body temperatures of 43° C and higher. This and other studies suggest the upper sustainable limit of subadult males aged 2-5 years lies between 40 and 43°C (Bartholomew and Wilke 1956; Irving et al. 1962; Ohata and Miller 1977a). I suspect the upper sustainable temperature of a pup lies between 40 and 41° C  Thermal Budgets & Climate Spaces  Chapter 5  72  and use a value of 41° C in subsequent calculations. Such values are comparable to the maximum core temperature taken from Galapagos fur seal pups of 39.8°C (Limberger et al. 1986). 5.4.3  M e t a b o l i c Rates  The thermal budget equation requires minimum and maximum metabolic rates in terms of the flux of heat to the pup's surface per unit area. In the case of fur seals, this means converting metabolic rates measured per unit of animal mass, taking into consideration the ability of pinnipeds to control bloodflowto the peripheral regions of their bodies. An important criterion in determining the extreme metabolic rates is that they be maintainable for some extended period of time. Blix et al. (1979a) found the oxygen consumption of a fasting pup, sleeping at air temperatures down to 6°C, was 3.5 W k g . If we consider the ability of the pup to - 1  increase blood flow to theflippersin response to heat (which increases the effective surface area by 40%), then the metabolic rate is approximately equal to 77 W m (i.e. 3.5 W k g  -1  - 2  x 5.4 kg / 0.2462 m ). I believe this to be the minimum sustainable 2  rate. Work by Worthy (1987) shows that the metabolic rates of harp seals and grey seals are lower (2.5—49.1%) during sleep than when resting. Therefore it is unlikely that a fur seal pup could maintain rates lower than the 3.5 W k g  - 1  recorded when  sleeping. Identifying the maximum sustainable metabolic rate (MR) is difficult. We know that the average daily metabolic rate of fasting pups (based on water influx values) is 7.56 W k g " for males and 10.00 W k g " for females (Costa and Gentry 1986). 1  1  We also know that values as high as 18 W k g  - 1  have been recorded from pups  immersed in ice water (Blix et al. 1979a), although it is unlikely that this 5-fold increase is sustainable. One experiment conducted on a 4-month-old female pup weighing 9.4 kg showed the resting metabolic rate at thermoneutrality was 0.72  Thermal Budgets & Climate Spaces  Chapter 5  mlg  -1  h  - 1  73  , while the maximum MR was 3.2 times higher at 2.3 m l g  W k g ] (Miller 1978). Young harbour seals -1  -1  h  - 1  [4-12.7  (Phoca vitulina richardsi) when worked  stressed through carrying lead weights while treading water, increased their oxygen consumption by a factor of 3.7 (Eisner and Ashwell-Erickson 1982). Studies of other mammals such as shrews (Campbell 1977), chipmunks (Heller and Gates 1971), and deer (Gates 1980) all further support the notion that the ratio of maximum to minimum sustainable rates lies between three and four. I have therefore assumed that the highest sustainable metabolic rate of a fur seal pup is 3.2 times more than its absolute minimum based on Miller's (1978) fur seal study. Thus if we accept the minimum MR is 3.5 W k g , then maximum MR would be 11.2 W k g . In terms - 1  - 1  of the amount of heat supplied to the pup's surface, the metabolic rate would be 409 W m  - 2  if the pup reduced its effective surface area by decreasing blood flow to  the flippers in response to cold [i.e. 11.2 W k g 5.4.4  - 1  X 5.4 kg / (0.2462 m x 60%)]. 2  Heat Transfer Resistance  The resistance to heat transfer of the coat and tissue (rnb) is determined in two parts. Thefirst,the resistance of the coat, was estimated from portions of fur sewn around a copper cylinder (Blix et al. 1979a). The thermal conductance of the pelt at birth was found to be 33.3 and 3.3 W m  - 2  °C  - 1  when wet and dry, respectively.  The second part, the tissue resistance, can be derived from measurements of heat flux density recorded from heat flow discs implanted subcutaneously in the backs of three newborn animals (Blix et al. 1979a). In dry air, heat flux through the skin was unrecordable (less than 100 W m  - 2  ) and subcutaneous temperature remained  within 1°C of rectal temperature. But, under a drizzle-like rain of 6°C, the heat flux rose to between 500 and 600 W m  - 2  and the subcutaneous temperature dropped by  20° C. This means the thermal conductance (I) of the wet pups was between 25 and 30 W m  - 2 o  C  - 1  , based on the relationship H = (T - T )I, b  8  where H is the  Thermal Budgets & Climate Spaces  Chapter 5  74  heat flux from inside the animal through a layer of fat or tissue, and Tj, and T are s  the temperatures of the body core and subcutaneous tissue (Gates 1980). Thermal conductance of the dry pup was about 100 W m sm  - 1  _ 2 o  C  _ 1  .  , the mean tissue resistance of a pup is about 44 s m  are constricted and about 12 s m  - 1  Converted to units of - 1  when the blood vessels  when they are dilated. The summation of the  tissue resistance and coat resistance (either 364 s m  - 1  if dry or 36 s m  - 1  if wet) is  equal to THb5.4.5  Surface Emissivity and Evaporative Heat Loss  Two parameters remain. The first, surface emissivity (e), has not been measured and is assumed to be 0.95, which is within the range for most natural surfaces of 0.90-0.98 (Campbell 1977). The second, total evaporative heat loss from far seals on land, is estimated from a study of California sea lions exposed to heat. In this sea lion study, evaporative heat loss accounted for 20% of heat production (Matsuura and Whitlow 1974). The same value has been shown for most homeotherms (Campbell 1977), which indicates total latent heat loss by respiration is around 20% of the metabolic rate (i.e. \E = 0.20M). 5.4.6  E n v i r o n m e n t a l Conditions  The thermal budget equation for Pribilof fur seals was solved for two levels of relative humidity (0% for dry pups and 100% for wet pups) and four wind speeds (u =0.1, 2.4, 7.4, 17.5 m s ) . The first wind speed corresponds to calm conditions and the _1  remaining three to the minimum, mean, and maximum daily wind speeds recorded during July from 1956 to 1981 on one of the Pribilof Islands, St. Paul (NOAA, National Climatic Center, Asheville, NC). Little or no weather data are recorded on San Miguel or any of the surrounding islands. Minimum and maximum average monthly air temperatures were available for only 3 years (DeLong 1982). Wind speeds and relative humidity levels were available from Los Angeles, which is 150  Thermal Budgets & Climate Spaces  Chapter 5  75  Table 5.1: Average daily weather conditions recorded at Saint Paul (1956-81) during July and at the Galapagos Islands (1982-84) during September. Wind speeds and relative humidity levels were unavailable for San Miguel Island and were hence taken from Los Angeles (1950-86), the closest weather station. The July air temperatures recorded at San Miguel are mean monthly values (1970, 1972-73). 1  2  3  4  Weather Condition Min. Wind Speed ( m s ) 2.2 Dry Air Temp. (°C) 4.4 Relative Humidity 0.6 -1  1 2 3 4  St. Paul Mean Max. 7.4 17.6 9.9 15.6 1.0 1.0  Min. 0.1 17.0  San Miguel Mean Max. 3.4 13.8 20.5 24.0 0.8  Galapagos Mean 1.17 23.8  -  NOAA, National Climatic Center, Asheville, N.C. Limberger et al. (1986) U.S. Dept. of Commerce (1987) DeLong (1982)  km east. Although this is the closest weather station to San Miguel, data recorded here should only be regarded as an index of possible conditions that the fur seals might experience on their rookeries. From 1951 to 1986, the minimum, mean, and maximum recorded wind speeds were 0.1, 3.4, and 13.8 m s  - 1  , respectively. The  range of air temperatures, relative humidities, and wind speeds for St. Paul and San Miguel are contained in Table 5.1. 5.4.7  T h e r m a l Budgets  The limiting lethal combinations of air temperature, wind speed, and radiation were calculated for the physical and physiological characteristics of northern fur seals outlined above. The surface area of a Galapagos fur seal pup weighing 3.4 kg is 1,809 cm according to the above method. The relationship between the mass 2  and length of northern fur seal fetuses (W = 1 0 - £ - 4  2  2 , 7 5  , Chapter 2) suggests the  pup is about 53 cm long. Bias in the diameter (6.5 cm) estimated from the above cylinder approximation was corrected by increasing the estimate by 12% to 7.3 cm. Metabolic rates for the Galapagos fur seal were extrapolated from the northern fur  Chapter 5  Thermal Budgets & Climate Spaces  76  Table 5.2: Parameters used to estimate the upper and lower physiological limits of northern fur seal pups and Galapagos fur seal pups. Metabolic rates are measured in W m , resistances are in s m , and temperature is in °C. - 2  - 1  Variable  n M XE dry r wet r r  Hb  r  Hb  Northern Fur Seal Thermal Thermal Minimum Maximum 34.0 41.0 409.0 77.0 81.8 15.4 408.0 376.0 80.0 48.0 192.2 179.6  Galapagos Fur Seal Thermal Thermal Minimum Maximum 34.0 41.0 351.0 65.8 70.2 13.2 408.0 376.0 80.0 48.0 192.2 179.6  seal studies. AH remaining model parameters were assumed equal to those of the northern fur seal pup, owing to a lack of data. Major model parameters for both species are found in Tables 5.2 and 5.3 for dry and wet pups (also see Table 5.4 for a listing of parameter definitions, values and units). Radiant E n e r g y Available  5.4.8  The minimum and maximum hradiances of the pup were estimated from Eqs. 5.2 and 5.3 as follows. Long wave absorptivity (a^,) of most plants and animals varies between 95% and 98% (Gates 1980) and was assumed to be 95% for the fur seal. Note that the seal absorptivity  is equal to the surface emissivity c (Kirchhoff's  Law). The average emissivity of the surroundings (e : clear sky and ground) was s  estimated at 0.77. Thus, with the inclusion of the Stephan-Boltzmann constant, the absorbed nighttime radiation calculated from Eq. 5.2 is 4.15 x 10 (273.15 + T ) _8  a  4  Wm" . 2  From Eq. 5.3, the maximum daytime irradiance from all sources on the Pribilofs at solar noon is 342.42 + 4.15 x 10~ (273.15 + T ) 8  a  4  W m " . This was calculated for 2  St. Paul Island (latitude 57°N) for July. The elevation angle of the sun from the  Thermal Budgets & Climate Spaces  Chapter 5  77  Table 5.3: Estimates of resistance (measured in sm ) used to determine the upper and lower physiological limits of fur seal pups. Columns numbered 1 to 4 correspond to wind speeds of 0.10, 2.24, 7.40, and 17.56 m s " on the Pribilof Islands and 0.10, 3.40, and 13.80 m s on San Miguel Island; and 1.17 m s on the Galapagos Islands. 1  1  - 1  Location  - 1  Variable  Pribilofs San Miguel Galapagos  r Ha r e  r  e  r  e  Thermal Minimum 1 2 3 4 286.4 60.5 33.3 21.6 115.0 46.0 28.4 19.4 286.4 49.1 24.4 115.0 39.1 21.6 76.7 54.8  Thermal Maximiam 1 2 3 4 286.4 60.5 33.3 21.6 110.4 45.3 28.1 19.3 286.4 49.1 24.4 110.4 38.6 21.5 76.7 53.7  -  horizon is 56.2 degrees at solar noon. This means that a pup oriented perpendicular to the sun will cast a shadow that is 26% of it's total surface area (A /A = 0.26). p  The calculated short-wave flux density (S  — 1103 W m  p  - 2  ) is a function of the  sun elevation angle and an atmospheric transmittance coefficient for a clear day of 0.84 (see Campbell 1977). The short-wave absorptivity of fur is set at 0.87 based on studies conducted on Galapagos fur seal pups (Limberger et al. 1986). Finally, mean diffuse solar irradiance (Sd = 107 W m ) was estimated using the methodology - 2  outlined by Campbell (1977). On San Miguel Island (latitude 34° N) the maximum radiation flux under full sun in July is 391.72 + 4.42 X 10" (273.15 + T f 8  a  W m " [from Eq. 5.3]. At this time 2  of the year, the elevation angle of the sun is 79 degrees at solar noon and northern fur seal pups cast shadows that are 30% of their surface area. Average emissivity of the surroundings is about 0.82. Further south on the Galapagos Islands (latitude 0°S), the sun's elevation is 81 degrees from the horizon during pupping in September and emissivity of the surroundings is 0.84. Approximately 30% of the surface area of Galapagos fur seal pups is exposed to direct sunlight (Limberger et al. 1986). Thus the maximum radiation flux on the Galapagos Islands, during September, is  Thermal Budgets & Climate Spaces  Chapter 5  78  Table 5.4: List of symbols and values.  Symbol a.  CiL A  Quantity short wave absorptivity long wave absorptivity surface area of pup  n.f.s. g.f.s.  1  2  A  p  A /A P  d  projected area on a plane perpendicular to solar beam fraction of surface area cast as shadow specific heat of air diameter of pup  n.f.s. g.f.s.  I  water vapor flux density heat flux thermal conductance  L  length of pup  M  metabolic rate parallel equivalent resistance of rjj and r radiative transfer resistance air boundary layer resistance coat and tissue layer resistance absorbed radiation diffuse solar irradiance  E H  r  e  a  r  r  THa THb  Rabt  s  d  a  b  W  surface emissivity of pup surface emissivity of surroundings  P PCp  air density heat capacity of air Stephan-Boltzmann constant latent heat of evaporization total evaporative heat toss  XE 1 2  air temperature deep body core temperature surface temperature of pup wind speed mass of pup  c  a A  = northern fur seal = Galapagos fur seal  wet dry n.f.s. g.f.s.  -  Units decimal fraction decimal fraction m m m  0.26 0.30 0.30  decimal fraction decimal fraction decimal fraction  -  Jkg~ C m m kg m s—1 Wm" Wm-^CWm-^Ccm cm Wm~ sm  2  2  2  l 0  0.087 0.073  -  - 2  -  2  25 100 62.0 53.0 Table 5.2 Table 5.3  1  1  2  - 1  T  short wave flux density perpendicular to beam T T T. u  St. Paul San Miguel Galapagos  Value 0.87 0.95 0.2462 0.1808  Table 5.2 Table 5.3 Table 5.2 St. Paul San Miguel Galapagos St. Paul San Miguel Galapagos  -  St. Paul San Miguel Galapagos  -  - 1  2 2 2  2  2  _ 1  -  0.20 M  - 1  2  Table 5.1 5.4 3.4 0.95 0.77 0.82 0.84 1200 5.67 x 1 0  - 1  2  106.9 108.6 108.7 1102.7 1138.7 1140.1 Table 5.1 Table 5.2  -  n.f.s. g.f.s.  sm sm sm Wm~ Wm" Wm" Wm" Wm" Wm~ Wm~ °C or °K °C or °K °C or °K m s kg kg decimal fraction decimal fraction decimal fraction decimal fraction kg m " J m " ' °K Wm" °KJ kg" Wm~ 8  1  -8  2  4  1  2  Chapter 5  Thermal Budgets & Climate Spaces  392.14 + 4.52 x 10~ (273.15 + T f 8  5.5  a  79  W m " [from Eq. 5.3]. 2  Results  Climate space diagrams are drawn for dry and wet pups on the Pribilof Islands (Fig. 5.1) and on San Miguel Island (Fig. 5.2). Within each polygonal space lie combinations of environmental conditions that are energetically tolerable by the pup. The space is bounded on the left and right by curves that represent the most extreme combinations of radiation and air temperature that could ever occur at each location. The left curve is defined by Eq. 5.2 and the right by Eq. 5.3. The upper and lower sets of lines represent the limiting combinations of air temperature, radiation, and wind speeds, above and below which the pup cannot survive (Eq. 5.1). The ends of all lines indicating physical and lethal limits are truncated to enhance the visual perception of the climate space. Pups can tolerate the combination of environmental conditions (air temperatures, radiation levels, and wind speeds) that he within the 'space'. The set of lines that describes the lower thermal limits of a dry pup includes extremely low air temperatures that are well below the most extreme conditions found on either island (top panels of Figs. 5.1 and 5.2).  Thus, only the upper  thermal limit is plotted for dry pups that are maintaining body temperatures of 41°C and a low metabolic rate of 77 W m  - 2  . These are plotted for pups subject to  four different wind speeds. The bottom panels of Figs. 5.1 and 5.2 show the climate space for a wet pup. Both upper and lower thermal limits are included for varying wind speeds. Note that the lower limits reflect a pup having a core temperature of 34°C and a high metabolic rate of 409 W m  - 2  .  Fig. 5.3 contains the climate space of a Galapagos fur seal pup weighing 3.4 kg and exposed to an average daily wind speed of 1.17 m s  - 1  as recorded by Limberger  et al. (1986). A second climate space of a northern fur seal pup exposed to the same  Thermal Budgets & Climate Spaces  Chapter 5  80  DRY  WET  0  200  400  600  800  Absorbed Radiation (W/m ) 2  Figure 5.1: Climate space diagrams for northern fur seal pups on the Pribilof Islands, Alaska. The polygons bound the combination of air temperatures, radiation levels and wind speeds that pups can tolerate when dry (top panel) and wet (bottom panel). The set of steady state lines describing the lower physiological limits of a dry pup are not shown because they are not contained within the scaling of the figure.  wind speed (1.17 m s ) on the Pribilofs is overlaid for comparison. - 1  5.6  Discussion  For the most part, our understanding of pinniped energetics has come from the laboratory, although more recently there has been an increasing emphasis placed on studies made under natural field conditions. While these studies have provided valuable insights, few have attempted to put all the pieces together into a mathematical model that considers the total thermal balance of pinnipeds (see Luecke et al. 1975; 0ritsland and Ronald 1978; Limberger et al. 1986). Porter and Gates's (1969) climate space analysis is one such approach to a holistic understanding of  Chapter 5  Thermal Budgets & Climate Spaces  81  DRY  WET  0  200  400  600  800  Absorbed Radiation (W/m ) 2  Figure 5.2:  Climate space diagrams for northern fur seal pups on San Miguel Island,  California. Rest of legend as in Fig. 5.1.  neonatal fur seal energetics and its environmental implications. The conclusions drawn from the linear thermal budget model are limited by the quality of the data used, which ranges from laboratory measurements to educated guesses. In each case I have tried to justify my choice of parameters and show my confidence in each estimate. Air temperature significantly affects heat transfer through animal fur (McClure and Porter 1983; Webb and McClure 1988) and may have been more extreme on the fur seal breeding beaches than I have considered. The standard height for meteorological measures is 2 m above ground. It is therefore possible that air temperatures in the south could be as much as 20° C higher at pup height during the day than the values recorded. In the north, ground surface temperature will tend to approximate sky temperatures at night and might be 10 - 15°C colder than local air temperatures.  Thermal Budgets & Climate Spaces  Chapter 5  60  -  82  DRY  40 20  £ 0) -20 0  -  ^ ^ ^ ^  V  k.  •2  -40  Co  1  t_  ///  •  1  ,  1  i  .  -  a>  WET  | - 60  — I  40  <  20 0  i  •  -  -  -20 1  0  .  1  200  .  1  i  400  .  i  600  800  Absorbed Radiation (W/m ) 2  Figure 5.3: Climate space diagrams of a northern fur seal pup on the Pribilof Islands (dashed lines) contrasted with that of a Galapagos fur seal pup on the Galapagos Islands (solid line). Both pups are subject to the same wind speed of 1.17 m s . Rest of legend as in Fig. 5.1. - 1  Unfortunately, better estimates of the microclimates at pup height are lacking. The results of a sensitivity analysis (Appendix 5.A) indicate that further refinement of the thermal budget parameters is unlikely to alter the general conclusions and predictions. Linearization of long wave radiation (Eq. 5.1) can introduce errors (Roughgarden et al. 1981; Tracy 1982); but they are probably small because the temperature of the fur air interface is unlikely to differ much from the free stream air. An additional error might arise from using a resistance model for fur bearing mammals because fur is a porous medium that does not behave like an ordinary solid insulation (Porter et al. 1986); however, such an error is likely minimal because the high hair density of fur seal pelts (Scheffer 1962) would cause conduction effects to predominate.  Chapter 5  5.6.1  Thermal Budgets & Climate Spaces  83  Weather and P u p M o r t a l i t y  The mortality of fur seal pups on the Pribilof beaches is often high and variable between years (Chapman 1961; Lander 1975, 1979b; Trites 1989). However, the results of the present model do not support the contention of Roppel et al. (1963) and Vladimirov (1974) that it is caused by inclement weather conditions. All the air temperatures, wind speeds, and humidity levels recorded on the Pribilofs (19561981) fall within a range that fur seal pups can tolerate (Table 5.1). Given the range of daily July air temperatures (4.44-15.65°C), the high humidity levels (0.65-1.00), and the general cloudiness of the region, pups appear to reside near the center of their climate space away from their lethal limits. Weather conditions would have to be more extreme than those recorded since 1956 to account for the high pup mortality observed on the Pribilof Islands. The ability of pups to cope with cold wet periods on the Pribilofs depends largely upon good cardiovascular control and maintaining high metabolic rates. Young fur seal pups are able to increase their heat production by both shivering and nonshivering thermogenesis (provided by loosely coupled mitochondria in certain muscles, Grav and Blix 1979). Body tissues and the thin fat layer provide pups with surprisingly little insulation. As for their fur, its good insulative properties are lost almost entirely when wet. Wind and rain compress the fur and reduce thermal resistance (Webb and King 1984). Note for example that a wet pup on a windy night would have trouble coping with air temperatures below 5°C (Fig. 5.1). Neither dry nor wet fur seal pups should have problems coping with high air temperatures at the Pribilof Islands. Air temperatures of 20 and 35°C are physiologically tolerable depending upon factors that dissipate heat from the body. Typical wind speeds at St. Paul Island (2.2-17.6 m s ) do not significantly affect the upper _1  lethal limit. However, the lethal limit changes markedly if the wind stops blow-  Chapter 5  Thermal Budgets & Climate Spaces  84  ing. According to Fig. 5.1, a dry pup exposed to full sun in still air (wind speed 0.1 m s ) , would die unless air temperatures stayed below -20° C. Wetting the pelt _1  would move the lethal limit up to about 5°C. It is clear that young pinnipeds, like all young mammals, have higher metabolic rates for their size than do similar-sized adults (Schmitz and Lavigne 1984; Lavigne et al. 1986). What is not clear is whether these high rates are due to the cost of forming new tissues while growing (Brody 1945; McNab 1980, 1986; Worthy 1987), or are the result of sustaining body temperatures during exposure to cold (Oftedal et al. 1987; Thompson et al. 1987; Whittow 1987). The possibility that weather could slow growth and thereby reduce the ability of fur seal pups to survive the rigours of the rookery and the subsequent winter migration, was not considered in the thermal budget. This model assumes that the high metabolic rate of lactating pups is maintained by unlimited supplies of lipid-rich milk. It does not account for interactions between nursing patterns and thermoregulation, affecting pup mortality, nor does it consider the tradeoff between growth and maintenance. Both these factors deserve further consideration and could enhance the predictive powers of the model. Unfortunately, few or no thermoregulatory data have been collected from fur seal pups fasted for prolonged periods of time. On San Miguel Island, fur seal pups are susceptible to heat stress. In contrast to conditions on the Pribilof Islands, radiation levels and air temperatures in California are higher and the wind speeds lower (Table 5.1). virtually no rain falls on San Miguel during the summer and sheltered air temperatures in excess of 21° C and surface sand temperatures in excess 42°C are recorded each year (DeLong 1982). Pups can contend with 20° C daily temperature if the wind is blowing, but to survive in still air must seek shade or wet their fur (Fig. 5.2). Part of the difficulty that northern fur seals have in the southern climate is due to their having no effective  Chapter 5  Thermal Budgets & Climate Spaces  85  way of losing heat by evaporation. 5.6.2  Behavioural Modifications  Pups should show a preference for an environment that is in the center of the climate space, away from the lethal limits. To achieve this, they can alter their metabolic rates and modify their behaviour. This latter option is the least costly. When cold, the seal should therefore move from wind to still air, huddle with other pups, curl up to reduce surface area, move into full sun, or remove water from the pelt by shaking and shuddering. A dry, heat-stressed pup should seek shade and breezes and wet its fur by entering the water. Field observations indicate that northern fur seal pups do seek shelter behind logs and rocks and huddle in groups during cold and windy weather (Ohata and Miller 1977b). In addition, pups frequently shiver in cold weather and shake water from their pelage when wet (Bartholomew and Wilke 1956; Irving et al. 1962). Similar behaviour has been noted for other otariid species (Whittow et al. 1971; Gentry 1972). Many species of otariids breeding near the equator are born on territories that are shaded and/or near the water. They protect themselves from extreme heating by keeping intermittently wet in the surf or in tidepools during sunny daylight hours. Such behaviour has been observed in the South American fur seal  (Arctocephalus  australis), Guadalupe fur seal (Arctocephalus townsendi), Galapagos fur seal, and Stellar sea lion (Peterson (Vaz-Ferreira and Palerm 1961; Peterson et al. 1968a; Odell 1974; Bonner 1984; Trillmich 1984; Fleischer 1987; Pierson 1987). Only one fur seal study has documented pup behaviour in detail. In it, Limberger et al. (1986) estimated that Galapagos fur seal pups were wet for an average of 3 h d  - 1  . They  also noted that lactating females usually dragged their pups with them to tidepools during the first few days after pupping.  Undoubtedly, this sort of behaviour is  critical to the survival of naive newborn pups. Similar behaviour appears to be  Chapter 5  Thermal Budgets & Climate Spaces  86  absent in the northern fur seal. DeLong (1982) observed the behaviour of heat-stressed northern fur seal pups on San Miguel Island after their mothers left for the splash zone. The abandoned newborn pups initially stayed in their territory, then began to move in a single random direction. All those that moved across the hot sand became prostrated, began convulsing, and died. Others moving towards the splash zone were revived by the cool water. Pups appeared to learn to move to the splash zone at about 10 days of age when the mother had left and the pup had joined other pups in exploratory movements away from their place of birth. 5.6.3  Role of B o d y Size  Another significant factor in the mortality of northern fur seal pups in southern climates is their large body size at birth (5.4 kg). This is shown by superimposing the climate space diagrams of the Galapagos fur seal pup and the Pribilof fur seal pup (Fig. 5.3). Inspecting the upper thermal limit indicates the lighter Galapagos fur seal pup (3.4 kg) can contend with air temperatures that are 7°C higher than the heavier northern fur seal pup, when both species are dry and exposed to the same wind speed. Body .size is not as important when the pups are wet. Under this condition the upper thermal limit of both species is almost identical. Thus the advantage of being heavy is only realized when animals are coping with the cold. The model makes several predictions about the expected pattern of mortality during inclement weather. The first is that heavier northern fur seal pups should succumb to the hot weather of San Miguel before smaller ones. Given that there is about a half kilogram difference in birth masses between the sexes, it follows that there is likely a greater mortality of males than females. The ability of fur seals to cope with high air temperatures will depend primarily on their ability to wet their fur. On the Pribilof Islands, where pups are brought close to their lower limits, the  Chapter 5  Thermal Budgets & Climate Spaces  87  model predicts a higher mortality of pups with low birth masses. There appears to be some general agreement between model predictions and the limited amount of data collected on pup mortality. On St. Paul Island, the birth masses and life histories of 410 pups were recorded (Calambokidis and Gentry 1985). Of the 25 pups that died from various causes, 10 were males and 15 females. Even more striking is the observation that all the dead pups were significantly lighter at birth than those in the total marked population. Perhaps small pups on the Pribilofs are more susceptible to sources of mortality because they must allocate a greater portion of their energy to maintenance than growth. On San Miguel Island there is some evidence that young male pups are more susceptible to heat than females. A 1981 survey of 106 dead pups (45 males and 61 females), found 98% of the males had died during hot weather compared with 71% of the females (DeLong et al. 1982). Most of the deaths (75%) occurred within 1 to 4 days postpartum. The model suggests that the climate of San Miguel Island should impose heavy mortalities on large pups, thereby suggesting strong selection pressures for pups with light birth weights. This sort of mechanism may account for the small size of the Galapagos fur seal (3.4 kg at birth). The Galapagos fur seal is closely related to the larger South American fur seal,  Arctocephalus australis (King 1983), which weighs  4.6 kg at birth (Vaz-Ferreira and Ponce de Leon 1987). Some have speculated that the ancestors of the Galapagos fur seal were South American fur seals that followed the northerly directed Peruvian current (King 1983; Bonner 1984). 5.6.4  Future Research  The proposed thermal budget describes the ability of the pup to cope with weather conditions during the first few days of life. A second critical stage in survival is  Chapter 5  Thermal Budgets & Climate Spaces  88  the transition from land to sea (Ichihara 1974), particularly in the Bering Sea. No marine mammals appear to have trouble in warm water (Whittow 1987). However, young fur seals seem to have difficulty maintaining their body temperatures at low sea temperatures, despite the development of adult pelage. The lower critical water temperature of a 4-month old northern fur seal pup was 16°C (pers. comm., L.K. Miller, University of Alaska, Fairbanks, Alaska). The pup had to shiver and raise its metabolic rate to maintain body temperature. Average sea surface temperatures of the Bering Sea are 7.38° C in October and drop to 5.45° C in November (Ingraham 1983). Thus, it should not be surprising if pups lose nearly half their body mass during their first winter at sea (Scheffer 1950, 1981). Given the severity of the pelagic transition, it would be useful to develop a thermal budget for this stage of life (Ohata et al. 1977). Several extensions of the pup's thermal budget might be of interest for future research. In addition to predicting thermoregulatory behaviour, the equation can be used to predict food requirements. The amount of milk required to maintain maximum sustainable metabolic rates can be determined by knowing the caloric value of the milk. Climate spaces could also be used to to gain insight into the worldwide geographic distribution of all pinniped species.  5.7  Summary  The ability of fur seal pups to cope with diverse climatic conditions on land was investigated by constructing a thermal budget based on published physiological studies. The model was applied to northern fur seals breeding on the Pribilof Islands, Alaska, and an San Miguel Island, California; also to the Galapagos fur seal pup on the Galapagos Islands. The combinations of environmental extremes that pups can withstand during the first week of life were identified. The study concluded that a healthy, average-sized pup on the Pribilofs could have tolerated any combination of  Thermal Budgets & Climate Spaces  Chapter 5  89  air temperature, wind speed, and level of humidity recorded since the mid 1950's, but that pups with low birth weights could have succumbed during periods of cold, wet, and windy weather. On San Miguel Island, the model predicts high mortalities of large pups during hot, dry weather, which suggests strong selection pressures towards the survival of smaller animals. The model further suggests the success of the Galapagos fur seal at the equator is related to its small body size and behavioural attributes, such as seeking shade and periodically wetting its fur.  5.A Appendix: Sensitivity Analysis The thermal budget described by Eq. 5.1 represents a linear relationship between air temperature (T ) and the amount of short- and long-wave radiation exchanged a  between the pup and the environment (Rabs)- It can be rewritten as  Rabs = a — b T  a  where  a =  f(E,T ,M,d,r ) b  Hb  T?rrA nr , = cxETt - c M + c ^  ~ 273.15)  - c  -j=  3  2  Mrffb 4  ^rrA - c*,Mr ETt Hb  and  with the constants a(i  = 1,...,5) set as follows: c\ = 5.67 x 10 ; -8  c  2  = 0.8;  c = 10.63; c = 7.09 x 10" ; and c = 1.51 x l O " . 3  3  4  10  5  The slope (b) is a function of the diameter of the pup while the intercept (a) is a function of five parameters (E,Tb,M,d, and rjjb). Each of these parameters is measured with errors eg, ej , ej^, ej, e . b  rm  If these uncertainties are independent  and random, then error in the estimated slope and intercept will equal  e a  =  [7da ndE E) e  V  Tda V JTa +{w >) {dM M) b  eT  +  e  \  2  (da {dd )  +  ed  \  2  TFa V {drVb ) +  erHb  Thermal Budgets & Climate Spaces  Chapter 5  90  Table 5.5: Parameter estimates (p±ep) and solutions of the partial differential equation (||) used to determine the total error (e ) in the intercept of the energy budget equation. See text for further details. a  p E  P .95 34.00 409.00 .087 80.00  n  M d e  P  eta  £l-oi#l 3.43 13.93 12.50 2.20 9.23 21.26  e  .02 1.00 40.00 .01 10.00  361.29 40.96 -3.06 -2524.85 -11.53  a  &  W.| 17.16 69.64 62.49 10.98 46.13 106.29  £1*1 7.23 40.96 122.22 32.82 115.32 176.20  and  Ifdb \  eb=  db db dd - e  2  i\dd )  ed =  d  This formulation identifies the contribution of individual parameter errors to overall model uncertainty (Taylor 1982). The thermal budget (Eq. 5.1) was applied to both dry and wet pups and considered upper and lower lethal limits at four different wind speeds, resulting in 16 unique solutions for the Pribilof Islands (see Fig. 5.1). Since the response of each model to errors is essentially identical I have chosen to present the sensitivity analysis conducted on the lower combination of air temperatures and radiation levels that are lethal to a wet pup exposed to a wind speed of 7.4 m s  - 1  (iZ b* = 412 — 36T ). a  a  Table 5.5 shows the estimates of the five parameters and the solutions of the partial differential equations used to estimate e . Because the actual errors associated a  with each parameter were unknown, I examined the effect of uniform errors of 1 and 5%. I also considered errors based on my insight into the maximum uncertainty that might be present for each parameter. If all five parameters were off by 1%, the intercept a would equal 412 ± 21 Wm  - 2  ; at 5% the error in a would be five times higher. If input errors are as high  as I have considered, then the uncertainty in the estimate of a could be as much as  Thermal Budgets & Climate Spaces  Chapter 5  176 W m  - 2  91  . Corresponding estimates of the error associated with the slope 6 are  36 ± 2 and 36 ± 10 W m  - 2  °C  - 1  for uniform errors of 1 and 5% respectively, and  36 ± 2 when considering maximum uncertainty. The intercept proved insensitive to errors in the diameter of the pup (d) and to errors in the estimate of surface emissivity (E).  The deep body core temperature  (Tj,), the amount of metabolic heat supplied to the surface (M), and the resistance of the fur seal coat and peripheral tissue layer  (rjjb) accounted for over 87% of the  uncertainty in a. Of these three parameters, I am most confident in Tb. This means the reliability of the model hinges upon the estimates of M and rubI feel that the possible error in the calculation of a and b are within acceptable bounds and that the predictions of the model are reliable. The method employed here is basically a summation of the effects of individual errors and is hence an indication of the maximum possible error that could occur.  The actual error in  a and b is likely much smaller than I have shown since introducing simultaneous errors into the thermal budget equation can compensate each other such that any new predictions would differ very little from those indicated in Fig. 5.1.  Chapter 6  The Influence of Weather on the Fur Seal's Life Cycle 6.1  Introduction  The arrival and departure of northern fur seals during the Pribilof breeding season is highly predictable (Bartholomew and Hoel 1953; Peterson 1965, 1968; Bigg 1986; Chapter 4). Territorial males begin hauling out in May to await the return of the females in late June (see Fig. 6.1). Most females arrive during the first two weeks of July and give birth within 24 h of being on land. Lactating females spend their first week ashore, then begin a regular attendance pattern consisting of 2 days nursing followed by a 4-8 day absence spent feeding at sea (Peterson 1965, 1968; Gentry 1981; Costa and Gentry 1986). The rigidness of the harem structure breaks down in late July, early August, as the territorial bulls begin leaving. Cows and pups do not leave until late October, early November. Peterson (1965) suggests that the precise timing of the annual life cycle is determined by the timing of molting or departure of the pups. He further states that the optimal time for birth is July because air temperatures are tolerable by pups and adults, and because pups have enough time to grow before being forced by storms to leave in November (Peterson 1968).  92  Chapter 6  Impact of Weather on the Life Cycle  800  93  -  May  Jun  Jul  Aug  Sep  Oct  Figure 6.1: Smoothed counts of territorial bulls, nursing females, and pups made by Peterson (1965) at Kitovi study site in 1962 and 1963. The data are representative of the annual cycle of fur seals on the Pribilof Islands. Ticks on the x-axis mark the first of each month.  My purpose is to determine if weather patterns in the Bering Sea are consistent with the annual synchronization of arrival, birth, and departure of Pribilof fur seals. The analysis summarizes weather data recorded on St. Paul Island from 1956-86. Monthly mean air temperatures, wind speeds, and relative humidities are contrasted with the timing of arrival and departure of fur seals to and from the Pribilof Islands as recorded by Peterson (1965). Results of the study are discussed in terms of the fur seals' ability to tolerate monthly weather conditions that typify the Pribilof Islands.  6.2  Methods  Weather data were collected by the Coast Guard Station on St. Paul Island and transcribed on magnetic tape. The weather was recorded every hour from 1956-63 and 1982-86, and once every three hours from 1964-81 (NOAA, National Climatic Center, Asheville, N.C.). Air temperature, wind speed, and relative humidity were selected for analysis because they are believed to have a major influence upon pup mortality (Bartholomew and Wilke 1956; Keyes 1965; Ohata and Miller 1976; Irving et al. 1962; Blix et al. 1979; Chapter 5). Monthly means and variances are calculated  Chapter 6  Impact of Weather on the Life Cycle  94  for the weather data and compared to the timing of pupping and migration. The time series data are further seasonally decomposed to separate inclement weather conditions from seasonal effects, using the method of Cleveland and Terpenning (1982) to interpret the complex patterns of weather. Seasonal weather patterns are analyzed and yearly trends and irregular weather conditions that might affect fur seals are noted.  6.3 6.3.1  Results and Discussion M e a n M o n t h l y Conditions  Average weather conditions (air temperatures, wind speeds, and relative humidities) are summarized by month in Fig. 6.2. The Pribilofs are consistently windy and wet, with mean monthly wind speeds varying between 7.5 and 8.5 m s  - 1  , and humidity  levels falling between 82 and 89%. The highest wind speeds are recorded from November to April when the fur seals are absent from the breeding islands. The arrival of territorial bulls in May corresponds to a sudden drop in average wind speeds and an increase in average air temperatures from 0.7° C in April to 4.2° C in May. Wind speeds drop to their lowest values in July when the pups are born. At the same time, air temperatures rise, peaking in July and August. 6.3.2  Seasonal Decomposition  Monthly time series of air temperatures, wind speeds, and relative humidities are seasonally decomposed in Figs. 6.3, 6.4, and 6.5. The top panel in each figure shows the monthly averages of the raw data sets. The data are decomposed into "a trend component that describes the long term variation in the series, a yearly seasonal component that describes variation that more or less repeats itself every year, and an irregular component that describes the remaining variation" not explained by the trend or the season (Cleveland and Terpenning 1982). The sum of the three  Chapter 6  Impact of Weather on the Life Cycle  95  T—i—i—i—i—i—i—i—i—i—i—r J F M A M J J A S O N D  Month Figure 6.2: Mean weather conditions (air temperature, wind speed, and relative humidity) recorded by month at St. Paul Island for the period 1956 to 1986. Vertical bars indicate 95confidence limits. components (trend, seasonal and irregular) equals the original time series. Removal of the periodic oscillations in weather conditions resulted in the seasonally adjusted time series (trend). The most notable alteration in weather trends was a decrease in humidity levels from 1956 to 1970, followed by a sudden increase in 1971. In general, air temperatures are much more stable and predictable than either wind speeds or relative humidity levels. Wind speeds appear to have fluctuated randomly since 1956. The only apparent trend has been a general decline in wind speed since 1974.  Impact of Weather on the Life Cycle  Chapter 6  96  Air Temperature 1956-86  52 CO  M  CO  u  a>  t  -4 • • i i i 1960  1965  1970  Year  1975  • 1980  i i i 1985  Figure 6.3: Monthly air temperatures recorded on St. Paul Island from 1956-68. The data are decomposed into a trend, seasonal effect, and irregular component (residual); the sum of which equals the original data set. Vertical bars at the right of each panel show the relative variation in scaling among the four plots.  Weather on the Pribilof Islands is dominated by a very strong seasonal component (Fig. 6.6). The highest air temperatures and relative humidities were recorded during the summer months of July and August, the same time that pups are born and nursed on land. The lowest monthly wind speeds were recorded during the months of June and July. It is equally apparent from the year to year evolution of the seasonal component (shown by the vertical bars in Fig. 6.6), that air temperature is much more stable than relative humidity and wind speed. This is also illustrated in the irregular plots (Fig. 6.7).  From 1956-86, there was virtually no  variation in air temperatures from May to October during the period of time that seals were ashore. This is unlike the irregular variation in wind speed and relative humidity which is consistently large throughout all months of the year.  Impact of Weather on the Life Cycle  Chapter 6  97  Wind Speed 1956-86 10  a„  S  8 6  9 c o K  9  8  7  0  J . J, J, ihi ill ill ill ill ,11 ,|| J l J . J  Ji Ji JJ i a m  1i l i i i J  V H r n r rTTi'TT vV if o ir in M r r f 11|j n  W -1 •  1960  I 1965  1970  Year  1975  I960  1985  Figure 6.4: Monthly wind speeds recorded on St. Paul Island. The data are decomposed into a trend, seasonal effect, and irregular component (residual). Vertical bars at the right of each panel show the relative variation in scaling among the four plots.  6.3.3  Impact of Weather on Seals  Northern fur seals seem to be born at the earliest point in time that will ensure optimal survival and time to grow. The Pribilof Islands are free of ice from May to December, although ice may rarely be present in June. Wind speeds in July, August and September are relatively low (respectively 7.1, 7.4, and 7.5 m s ) and - 1  air temperatures are their warmest (9.5, 10.2., and 8.3°C respectively); thus ensuring the pup three months of tolerable weather (see Chapter 5). The major difference between June and July is the change in mean air temperature from 6.8 to 9.5° C. If pups were born in June, they would experience low air temperatures, in combination with the general high winds and humidity of the region, which would likely result in high mortalities (Chapter 5). Thus it seems that the critical factor determining the survival of pups and hence the mean date of birth is air temperature.  Impact of Weather on the Life Cycle  Chapter 6  98  Relative Humidity 1956-86 Transformation Power -1 -1.00  -  1960  1965  1970  1975  1980  1985  Year Figure 6.5: Monthly levels of relative humidity recorded on St. Paul Island from 195686. The data are transformed according to x' = x~ and are decomposed into a trend, seasonal effect, and irregular component (residual). Vertical bars at the right of each panel show the relative variation in scaling among the four plots. x  Pups develop adult-type underfur and guard hairs in mid August, prior to the September molt (Scheffer 1962). Concurrently, there is a drastic deterioration in weather conditions from September to November as air temperatures drop (from 8.7 to 1.4°C) and winds increase (from 7.6 to 8.4 m s ) . The severity of the November -1  weather is presumably intolerable by both the pups and their mothers, and is thus the limiting factor determining their length of stay on the Pribilof Islands. This view is consistent with field observations indicating pups and cows begin leaving just prior to November 1 and are entirely absent from the Islands by November 20 (Fig. 6.1). The fur seal's life cycle may revolve about the timing of birth and the survival of nursing of pups rather than the timing of molting and departure of pups as suggested  Chapter 6  Impact of Weather on the Life Cycle  T J  1 1 1 1—i F M A M J  1—i—i J A S  99  1—i—r O N D  Month  Figure 6.6: Seasonal component of air temperature, wind speeds and relative humidity levels from 1956-86. Horizontal lines are at the midmean of the seasonal component for each month; vertical lines show year-to-year evolution of the seasonal component. Note that relative humidity is transformed according to x' = x ' . 1  Chapter 6  Impact of Weather on the Life Cycle  100  Figure 6.7: The irregular component of air temperatures, wind speeds and relative humidity levels from 1956-86. Horizontal lines are at the midmean of the irregular component for each month; vertical lines show year-to-year evolution of the irregular component. Note that relative humidity is transformed according to x' = a ; . - 1  Chapter 6  Impact of Weather on the Life Cycle  101  Table 6.1: The average surface temperature of the Bering Sea by month. The mean temperatures were calculated from available data spanning the period 1953 to 1982 (Ingraham 1983) for the continental shelf area surrounding the Pribilof Islands.  Month January February March April May June July August September October November December  Surface Temperature (°C) 2.13 1.46 1.43 2.09 3.32 5.65 8.20 9.41 8.90 7.38 5.45 3.54  by Peterson (1965). Weather conditions in July, when pups are born, are optimal for pup survival, but would be unacceptable in June (Chapter 5). Weather conditions continue to remain favorable for the nursing pups during August and September. Air temperatures vary little within any given month, from one year to the next, and may be the critical factor influencing the survival of newborns (Chapter 5). Annual weather conditions on the Pribilof Islands are consistent with the synchronized timing of pupping, and may be the major influence in the timing of the fur seal's life cycle. But such a conclusion fails to recognize an equally critical factor, food, which is suggested to be the prime determinant of the precisely timed cycles of elephant seals (Carrick et al. 1962b) and is undoubtedly a key component of the annual cycles of fur seals as well. There is no clear picture of the seasonal abundance of fur seal prey species about the Pribilof Islands. There are some suggestions that food abundance can be inferred from the influence of water temperature upon fish distributions (Alton 1974; Walsh and McRoy 1978; Mclain and Favorite 1976). Fur-  Chapter 6  Impact of Weather on the Life Cycle  102  thermore, a high correlation exists between Bering Sea water temperatures (Table 6.1) and monthly air temperatures (r =80.2%, t=6.76). However, the possibility 2  that optimal weather conditions and food occur simultaneously will require further study to be properly evaluated.  6.4  Summary  Weather conditions recorded from 1956-86 on St. Paul Island, Alaska, are probed to establish their effect upon the northern fur seal's life cycle  (Callorhinus ursinus).  Air temperatures, wind speeds, and relative humidity levels are seasonally decomposed and compared with the timing of pupping and migration. The seal's life cycle appears to revolve about the timing of birth and survival of nursing pups. Most pups are born in early July when air temperatures and relative humidity are approaching their highest annual levels and wind speeds are at their lowest. Weather conditions are optimal for growth and survival of pups from July to September, but are unacceptable in June. A rapid deterioration in weather through October and November corresponds with the fall migration of pups and lactating females. Air temperatures are the most dominant seasonal component and may be the proximate factor influencing the precise timing of the northern fur seal's life cycle.  Chapter 7  Does Tagging and Handling Affect the Growth of Pups? 7.1  Introduction  In 1957, northern fur seal pups on the Pribilof Islands were weighed a few weeks after tagging and found to be lighter than untagged pups (Abegglen et al. 1957). Over the next 9 years biologists continued to report that untagged pups outweighed tagged pups (Roppel 1984). Their conclusion was "tagging, marking, and handling, individually or combined, causes a loss of weight or slows the normal rate of weight gain. Loss of weight may cause tagged or marked pups to die at a greater rate than untagged and unmarked pups during their first winter at sea, thus inflating later estimates of the population based on recoveries of tagged and marked seals."  1  Research into the effect of tagging and handling on pup growth ended in 1966 as interest in tagging seals waned. Today, Pribilof biologists are expressing a renewed interest in tagging (see Fowler 1986). The possibility that handling pups might slow growth and reduce future survival has implications for pinniped studies that rely on marking individuals for future identification. It also has bearing upon the interpretation of historical data collected from tagged northern fur seals and is relevant to understanding whether the mark-recapture method for estimating the 'from M M B L (1969)  103  Chapter 7  Does Tagging Affect Pup Growth?  104  number of pups born is detrimental to the survival of the Pribilof population. Could the shearing of thousands of pup heads (see York and Kozloff 1987), be reducing the young animals' growth rates and future survival? The goal of this chapter is to re-assess the data collected on tagged and untagged pups from 1957-65 to determine whether tagging affects pup growth. I also reexamine data collected during 1965-66 from pups that were marked but not tagged, to see whether the type of marking is related to pup growth. I begin by reviewing the methods used to tag and later weigh the pups. I then compare the different masses of tagged and marked pups, and test whether tagged pups actually grew at a slower rate than untagged pups.  7.2 7.2.1  Materials and Methods Tagging, M a r k i n g & Weighing  Fur seal pups are born from the last week of June until the end of July (Bartholomew and Hoel 1953; Peterson 1968). The number of births over time is positively skewed with over 50% occurring during the first two weeks of July (Chapter 4). On average, the pups were about 5-6 weeks old when tagged in mid August, and 7-8 weeks old when weighed from Aug 29 - Sep 3. From 1957-65, monel metal cattle-ear tags were clinched at the hair line on the front nipper (1957-64) or between the fourth and the fifth digit (1964-65). Checkmarks, used to identify pups that lose the tag, were made by cutting a V-notch near the tip of a flipper or by slicing off a flipper tip. Tags were allotted to each rookery according to the proportion of harem bulls counted on that rookery. From 1957-62, approximately 10 000 pups were tagged at each of Northeast Point and Reef rookeries (see Fig. 12.1 for locations). Fewer pups were tagged at Polovina (4 400) and Zapadni Reef (3 500). The number of tags attached prior to weighing dropped from 1963-65 (2 400 Reef; 2500 Northeast Point; 1000 Polovina; 900 Zapadni Reef).  Chapter 7  Does Tagging Affect Pup Growth?  105  Pups were rounded up for tagging and herded towards barricades. They were lifted onto tables, tagged, and released seaward of the barricades. Some areas were covered more than once to achieve the desired number of tagged seals. Information on the effect of handling as opposed to tagging was obtained in two years. In 1965, some untagged pups were marked by removing the tip of the first digit on the right hind flipper and weighed 12-14 days later (Roppel et al. 1966). In 1966, 800 pups on Zapadni Reef and an equal number at Northeast Point were marked by shearing a patch of fur from the rump. They were weighed 13 days after marking (MMBL 1969). Pups were weighed on a spring scale in the early years, and later on a platform scale (Roppel et al. 1966). While the technique for weighing the pups was refined over the years, it remained consistent within years such that tagged and untagged pups were weighed under identical conditions in any given year. For weighing, pups were rounded up and herded towards barricades. Pups were chosen from the 'mob' that ensued, were weighed, and were released beyond the enclosure to prevent re-weighing. From 1957-61, the annual number of tag bearing pups weighed on a rookery varied between 1 and 111 of each sex, and averaged about 50. The number of untagged pups also varied during the same period and averaged about 100 animals of each sex. From 1962-66, biologists annually weighed 75 tagged and 75 untagged pups of each sex. 7.2.2  Analysis of G r o w t h  Patterns  The mean mass of male and female pups (untagged, marked, and tagged) are estimated for each rookery and year when more than 30 pups were weighed. Differences between the mean mass of tagged and untagged pups are then determined and 95% confidence intervals calculated. I also re-examine the sizes of marked, tagged, and untagged pups recorded in 1965, and test, by analysis of variance, whether changes  Chapter 7  Does Tagging Affect Pup Growth?  106  in size are related to the type of mark applied. The effect of tagging on growth is considered in three ways. First, the mean mass of tagged pups is compared with the mean mass of untagged ones. Next I plot the difference in size  ([untagged — tagged]/untagged X 100%) against the number  of days elapsed since tagging to see whether a relationship can be attributed to the effects of tagging. Finally, I construct growth curves for tagged and untagged pups and contrast them with one another. Growth curves are drawn for two different time periods. The first is a re-analysis of three weighings of tagged and untagged pups made a month apart in 1962, that began one week after tagging (Roppel et al. 1963). The second growth curve pools the mean mass of pups recorded on each rookery from 1958-66 and plots them against the day they were made (between Aug 29 - Sep 3). I fit separate linear regressions to the tagged and untagged data sets, and determine the equality of the two population regression coefficients with a Student i-test (Zar 1984). If the slopes (i.e. growth rates) are not statistically different I then use a second t-test to test whether the elevation of the two growth curves is also the same.  7.3  Results and Discussion  With few exceptions, the mean mass of pups not previously handled or tagged consistently exceeds the mean mass of tagged and marked individuals on all rookeries (Figs. 7.1 and 7.2). In some years the mean difference between the two groups is as much as 1.7 kg. There is no consistent pattern in the mean difference on the different rookeries. Considerable variability exists in the mean difference among sexes. In some years, and on some rookeries, the mean difference between the size of tagged and untagged pups is the same for both males and females. At other times and places (e.g. Reef 1957, Fig. 7.1) there is a marked difference between the apparent effect of tagging on males and females.  Does Tagging Affect Pup Growth?  Chapter 7  • Male Pups  CD _  3  -  2  -  1  -  — "D 0 e oS >  Q to  -1  C  CO  co c CD O  2  NEP  II  -  w ,  !*  ° Female  ZRF  83  CD *CD *= —  107  POL 2  -  1  -  REF  0 -1  -  T—r  T—I—r57  59  61  63  T  T—i—i—i—i—i—i—r  65  57  61  59  63  65  Year Figure 7.1: Annual differences in the mean body mass of tagged and untagged, male (circles) and female (squares) pups weighed on four rookeries: Zapadni Reef (ZRF), Northeast Point (NEP), Polovina (POL) and Reef (REF). Most of the points lie above the dashed line of no difference, indicating untagged pups weighed more than tagged pups. Solid lines are 95% confidence intervals for the mean difference.  Samples of marked and tagged pups weighed in 1965 are smaller than those of previously unhandled pups (Fig. 7.3). But, there does not appear to be a significant difference between the mean mass of marked pups and tagged pups. The size of marked and tagged female pups does not differ on each of the four rookeries 1.767, p = 0.184).  (F\^s& =  The difference in male mass (Fi, 8 = 4.039, p = 0.045) is 58  attributed to the Zapadni Reef sample. There is no significant difference in the size of marked and tagged males on the other three rookeries (^1,441 = 1.122, p = 0.290).  Thus I conclude that the type of mark (either tagging or slicing the flipper tip) does not affect the size of the pup, and will, for the remainder of this chapter, refer to all  Chapter 7  Does Tagging Affect Pup Growth?  108  r-0.746 p<0.001  Male Pups - a - Female Pups  • v •  CO  r-0.683  o  p<0.001  — I 6 65  7  8  9  10  11 1  Untagged Mean Mass (kg) Figure 7.2: Mean mass of tagged pups versus those without tags. Each point represents one rookery and one year, where the mean mass was calculated for more than 30 pups. The significance of the linear regressions is shown at the top and bottom of the panel for males and females, respectively. The different elevation of the parallel regressions shows the size difference between males and females.  marked and tagged pups as being tagged. 7.3.1  G r o w t h Rates  In 1962, tagged and untagged pups were weighed on three occasions: Sep 2-3, Oct 2-3 and Oct 24-25 (Fig. 7.4). Roppel  et al. (1963) applied a t-test to each pair of  observations according to date and rookery. They concluded that weight differences within rookeries were significant at the first weighing but were insignificant at the third weighing. In other words, they concluded the immediate weight loss caused by tagging was partially overcome after 2 months. The lack of significance between the weights of tagged and untagged pups at the third weighing is partly explained by the small sample sizes and by the large variation in body size attributed to growth. Pooled estimates of the standard deviation of male mass increased between the successive weighing periods from 2.00 to 2.46, to 2.98 kg. The standard deviation of female mass on the three occasions was 1.64, 2.12, and 2.41 kg. Sample sizes of 75 were sufficient on the first day of weighing  Chapter 7  Does Tagging Affect Pup Growth?  Zapadni Reef  Reef  Polovina  •  Untagged & Unmarked  •  Tagged & Marked'  0  Marked  109  Northeast Point  Rookery Figure 7.3: Mean mass of the tagged, marked, and untagged/unmarked pups weighed on four rookeries on Sep 2-3, 1962. Untagged pups were always heavier than those tagged and marked. Similarly, males weighed more than females. The solid lines show the 95% confidence intervals for the mean.  to detect differences as small as 0.75 kg between the mass of tagged and untagged females, but should have been increased to 112 and 144 on the second and third weighings to maintain the power of the t-test (when a = 0.05, /? = 0.25). The sample size necessary to detect the same mass difference among males should have been 95, 150 and 220 pups for each successive weighing. Thus it cannot be concluded with assurance that tagged pups regained their weight loss 2 months after tagging, because the probability of detecting a true difference among the mean population mass of tagged and untagged pups on the third weighing was too small (i.e. j3 was greater than 0.80). Fitting linear regressions to the 1962 data, pooled from all four rookeries, suggests that tagged pups grew at the same rate as untagged pups over the periods Sep  Does Tagging Affect Pup Growth?  Chapter 7  15  Males I  r-0.927 (K0.001  13  -B- Untagged  11  D) (0 CO  110  //  »//  °  — - Tagged  is •  9  7  r=0.902 (XO.001  5  Q. 15 Q. c 13 CO CD  Females r.0.906 (X0.001  O  11 9 a  7  s  • r=0.951 p<0.001  5 1  15 31  Jul  Aug  Sep  Oct  Nov  Figure 7.4: Mean body mass of males and female pups measured on four rookeries from September to October 1962. Linear regressions describe two periods of growth for tagged and untagged pups. The significance of the separate regressions are contained in Table 7.1, while the significance of a single regression of the mass of pups over the three time periods is shown in the tops and bottoms of each panel respectively.  2-Oct 3 and Oct 2-24, but were smaller than average when tagged (Fig. 7.4 and Table 7.1). The same conclusion is drawn from growth curves constructed for the six day period, Aug 29-Sep 3 (Fig. 7.5 and Table 7.1). It would seem that both groups of pups started out in time at different sizes. Unlike the mass of tagged males, tagged females, and untagged females shown in Fig. 7.5, the linear regression of untagged male mass from Aug 29-Sep 3 was not significant because of the large variance associated with the larger male body size. A significant relationship occurs when the number of data points is increased by including untagged males weighed in 1967-71, 1984 and 1987 ( F  li55  = 4.97, p - 0.030,  Does Tagging Affect Pup Growth?  Chapter 7  Ill  Table 7.1: Linear regression coefficients (growth rates) estimated for tagged and untagged pups weighed in 1962 and 1957-66. The equality of the two population regression coefficients was tested with a Student's t (tslope)- A second Mest was used to test whether the elevations of the two population regressions were the same (t Uvation )• The probabilities associated with the sample statistics are enclosed in brackets. The rate of growth from 1957-66 was determined using the annual mean estimates of mass from each rookery where at least 30 pups were weighed. e  Time Period year days 1957-66 Aug 29 - Sep 3 1962  Sep 2 - Oct 3  1962  Oct 2 - Oct 25  Sex M F M F M F  Growth Rate (kg d ) tagged untagged 0.179 0.131 0.180 0.160 0.093 0.092 0.081 0.077 0.057 0.046 0.073 0.061 - 1  (P)  0.432 0.214 0.120 0.544 0.742 0.952  (0.334) (0.416) (0.452) (0.293) (0.229) (0.171)  ^elevation  (P)  -4.307 (<0.001) -4.005 (<0.001) -5.768 (<0.001) -4.864 (<0.001) -3.547 (<0.001) -2.517 (0.006)  sample size 56 58 1191 1194 1191 1196  see Chapter 8). The slope of this regression (the growth rate) was not significantly different from the slope for tagged pups (ts3 = 0.486, p = 0.314), but regression elevations did differ (ts3 = -5.974, p < 0.001). Growth rate of males and females slows from August to November, the moulting period. Untagged males went from 0.131 to 0.092 to 0.046 kg d  - 1  as the season  progressed (Table 7.1). The reduced growth rate may reflect a decline in the amount of energy that a mother can transfer through her milk to her growing pup, or it may be that pup metabolism increases more rapidly with size than does feeding rate. The duration of female feeding trips at sea (absence from the rookery) is known to increase over the breeding season, and is believed to reflect the nutritional needs of the pup (Gentry and Holt 1986). Presumably milk cannot meet the needs of a growing pup for more than about four months; the pup is forced to increase its energy intake with solid foods. Curiously the growth rates of tagged pups, although not statistically different, always exceeded the growth rate of untagged pups (Table 7.1). Perhaps tagged pups were born later than untagged pups and hence were in a different growth phase or  Does Tagging Affect Pup Growth?  Chapter 7  12 -  Males r-0.293 p-0.122  10  D)  112  - a - Untagged - • - Tagged  8  vs CO CO  ^  D_  cf c  6 1 2  Females r-0.427 p-0.019  CO  0 10 8 -  6 -  Figure 7.5: Mean body mass of tagged and untagged pups weighed from Aug 29 to Sep 3. Each point represents one rookery and one year from 1957-66, where the sample size was greater than 30. The linear regressions were fit to the mean data points rather than the raw values to reduce the combined effect of annual differences in body sizes and sample sizes. The significance of the regressions is shown in the top and bottom of the panels.  perhaps the growth curve of tagged pups lagged behind that of untagged pups. Or both. If tagging did indeed affect growth, then there should be a relationship between the size of the pup and the number of days elapsed since tagging (Fig. 7.6), but there is none. Thus it must be concluded, based on this and the above evidence, that tagging and handling did not affect pup growth, but that the pups selected for tagging were smaller than average and hence were not representative of the whole population. Conceivably, the pups rounded up for tagging were born late in the season.  Perhaps many pups born earlier were in the surf and tide pools at the  Chapter 7  Does Tagging Affect Pup Growth?  113  • •  O  •  I a A  •  Male Pups  •  Female Pups  •  a  0 5 10 15 20 Days Between Tagging & Weighing Figure 7.6: Differences between the mean body size of tagged and untagged pups subsequent to tagging. Each data point shows how much larger untagged pups were compared to tagged individuals (expressed as a percent of tagged body mass). Each value represents the mean of one rookery and year from 1958-66. The dates of tagging were not available for 1957. There appears to be no relationship between the effect of tagging and the number of days elapsed since the tag was attached.  time of tagging and could not be rounded up. Or perhaps pregnant females that returned to land later in the breeding season were forced to give birth in the lowdensity peripheral regions of the rookery, such that their pups were more susceptible to being captured. Furthermore, taggers may not have chosen pups at random from the numbers they rounded up. Pups tend to pile up (and even smother) when surrounded. Perhaps smaller pups were taken from the top of the piles, leaving the heavier pups at the bottom. A more detailed discussion of these possible sources of biases is contained in Chapter 8. 7.3.2  O t h e r Species  Tagging pinnipeds facilitates studies of life history, population biology, behaviour, growth and development, dispersal from rookeries and fidelity to birth sites. It also permits validation of teeth-aging techniques (Condy and Bester 1975; Summers and Witthames 1978). Despite the tagging of many species of fur seals (e.g. Rand 1950,  Chapter 7  Does Tagging Affect Pup Growth?  114  1959; Csordas and Ingham 1965; Bonner 1968; Crawley and Brown 1971; Stirling 1971b; Laws 1973; Condy and Bester 1975; Payne 1979; Mattlin 1981; Kerley 1985), there is little or no information about the possible effect that tagging might have on pup growth. Observations of Antarctic fur seals,  Arctocephalus gazella, led Payne  (1979) to suggest that pups tagged within one month of birth grew slower than untagged pups, although no data were presented. Further studies of Antarctic fur seal pups, tagged when two months old, showed no abnormal growth rates (Kerley 1985), nor did tagged New Zealand fur seal pups,  Arctocephalus forsteri (Mattlin  1978, 1981). Possibly the tags did not alter growth because they were attached at a less critical time in the development of the pup (Kerley 1985). On the other hand, it should be noted that none of the authors presented their data, and that their conclusions appear to be based on small sample sizes. Thus it is doubtful that differences between the mean size of tagged and untagged pups could have been detected had they actually been present. Results from reassessing pup mass data from northern fur seals suggest that tagging too early does not alter growth rates, but results in selecting individuals that are smaller than average. Possibly, pups of different ages are not randomly distributed on the rookery during the first month of pupping, but are segregated by size, such that small pups' are more easily accessible for tagging and weighing than are larger ones. My conclusion is that tagging and handling do not significantly affect pup growth.  This opinion could be strengthened or weakened by weighing pups at  the time of tagging, recording their tag numbers, then resampling at later dates to compare changes in weight. This procedure would undoubtedly reveal some of the possible biases inherent in the sampling design and shed light on the biological factors that cause differences in average mass. Researchers must be aware of the large variances in mass and the difficulty of recapturing marked animals and therefore take statistically large samples to detect any changes in body size. This  Chapter 7  Does Tagging Affect Pup Growth?  115  kind of research should be done before committing resources to long-term tagging programs.  7.4  Summary  From 1957 to 1966 samples of tagged and marked pups consistently weighed less than untagged and unmarked northern fur seals. At the time, it was concluded that tagging and handling had caused a loss of weight and had slowed the normal rate of pup growth. Li re-evaluating the data from this time period, it seems that tagged pups grew at the same rate as untagged pups, but were smaller at the time of tagging than average size pups. The growth curve for tagged pups appears to lag behind that of untagged pups, suggesting that tagged pups were born later in the breeding season and were more susceptible to being captured and tagged than older and heavier pups.  Chapter 8  Biased Estimates of Pup Mass: Origins and Implications 8.1  Introduction  Pup weights axe relatively easy to obtain and are available for all species of fur seals (King 1983). Pups have been weighed over successive days to construct growth curves for northern fur seals (Scheffer and Wilke 1953), Subantarctic fur seals,  Arc-  tocephalus tropicalis (Kerley 1985), Antarctic fur seals, A. gazella (Doidge et al. 1984a; Kerley 1985), Galapagos fur seals, 1985) and New Zealand fur seals, as the Cape fur seal,  A. galapagoensis (Trillmich and Limberger  A. forsteri (Mattlin 1981). For other species, such  A. pusillus pusillus (Rand 1956), only a single set of descriptive  pup weights are available. Growth curves have been contrasted with one another to make inferences about life history strategies of the different fur seal species, and have been used to compare the rate of pup growth of a single species between years (e.g. Kerley 1985). Annual differences in the mean size of pups are potentially a valuable management tool if, as believed, they reflect maternal competition for food and influence the future survival of the young animals (Scheffer 1955; Chapman 1961; Eberhaxdt and Siniff 1977; Mattlin 1981; Doidge et al. 1984a; Kozloff and Briggs 1986; Doidge and Croxall 1989).  116  Chapter 8  Biased Estimates of Pup Mass  117  In order to properly construct growth curves or compare the size of fur seal pups born in different years, several factors, such as when the pups were weighed and how large the samples were, must be carefully considered and understood. Depending upon the rate of growth, it may be invalid to compare pups weighed on different days in different years. Similarly, difficulties may arise if samples contain primarily fasted pups or consist of pups that have recently fed. Finally biases associated with sample sizes, behavioural segregation of different aged pups, and human error in randomly selecting pups for weighing need evaluation. In this chapter, I evaluate the potential problems associated with weighing fur seal pups. I focus on three areas: the mass of milk consumed, the rate of growth during the sampling period, and the effect of different sample sizes. My goal is to document each factor individually and provide an overview of the problems to consider when weighing fur seal pups.  8.2  Milk Consumption  I derived the mass of milk consumed from the work of Costa and Gentry (1986). They measured total water influx and determined that pups, on average, consumed a total of 3.5 liters of milk during the first 7 days following birth and 3.4 liters during each subsequent 2 day feeding bouts. Their report contains twofiguresshowing the volume of milk consumed per feeding bout plotted against (1) the age and (2) the mass of the pup. Digitizing this data, I produced a third figure (body mass versus age) and estimated the volume of milk consumed per day of feeding. I then converted volume to mass by multiplying the components of milk by their specific densities (Table 8.1). Fur seal milk consists of 44.4% water, 41.5% fat, 14.2% protein, and 0.5% ash (Costa and Gentry 1986); and has a density of 1.021 (Table 8.1). Given that the mass of water at 28°C is 0.9982 g m l  - 1  (Kleiber 1961), the mass of fur seal milk  Biased Estimates of Pup Mass  Chapter 8  Table 8.1:  118  Estimated density of northern fur seal milk determined from the specific  density (Kleiber 1961) and percent composition of fur seal milk components (Costa and  Gentry 1986). Components  Mean Composition 44.4% 41.5% 14.2% 0.5%  water fat protein ash lactose total is 1.019 kg l  -  Density 1.00 0.89 1.35 3.10 1.52  Density of Milk 0.444 0.369 0.192 0.016  -  1.021  - 1  (calculated as 0.9982 x 1.021). This estimate of density assumes  that fur seal milk contains no carbohydrates (lactose) which is in keeping with the conclusions of Pilson and Kelly (1962) and Pilson (1965). However, Schmidt et al. (1971) found low, but detectable lactose concentrations that suggest the above calculations are slightly underestimated, although not significantly so. The mass of milk consumed in one day of feeding varied between 0.2 and 3.8 kg d  - 1  depending upon a pup's size and sex (Fig. 8.1). It seems that males con-  sumed more milk than females because they were physically larger. Note also that consumption increased as the pups grew over time (cf. Costa and Gentry 1986). Although the amount of milk consumed is shown in Fig. 8.1 as a linear function of time and body mass, it is more likely a nonlinear relationship which approaches an asymptote. Presumably, the capacity of lactating females to produce milk is limited at some point in time due to physiological constraints or the process of weaning. Milk accounts for a large proportion of total body mass. For example, an average sized male pup weighing 9 kg before feeding, is likely to consume between 3.0 and 7.5 kg of milk over a two day feeding bout (Fig. 8.1). Thus the pup's mass could vary considerably depending upon whether the pup is weighed at the end of the  Chapter 8  Biased Estimates of Pup Mass  119  12  Ma SS  ( M l  Q_  Z> Q_  10  -  8  -  6 4 4  Ma  CO  ii  Male Pups -a- Female Pups  * •-. •  • o *> '  -  n " '  o,* O  a  o  ...  r=0.867 (X0.001  _ (•0.728 (X0.001  (=0 6S1 |x0.001  D)  '—'  001  •  •  3  •  2 .oo •„ - -  1  T-  -  *"  a o (•0.894  e  0 i  i  - <  r  1  1  1  1  1  .  o°  jxO.001  30 10 20 30 10 20 30 10 Jul Aug Sep  8  1  1  4  6  1  8  1  1  10  12  Pup Mass (kg)  Figure 8.1: Growth and milk consumption for male and female pups during one day of feeding. Males were significantly larger than females (top left panel) and consumed greater amounts of milk (bottom left panel). The greater consumption of milk by males is related to their larger body size. Linear regressions applied to milk mass as a function of body size for each sex separately were not significantly different (slope: <45 = 1.119, p = 0.135; elevation: £15 = 1.761, p = 0.043), indicating that males and females of comparable body sizes consumed the same amount of milk (bottom right panel). A linear regression is therefore shown for both sexes combined. The significance of the regressions is contained within each panel. The data shown in this figure were derived from the work of Costa and Gentry (1986) were assumed to be from pups born on July 7, the mean date of birth (Chapter 4).  Chapter 8  Biased Estimates of Pup Mass  120  5 day fast or immediately after the 2 day suckling period. Such a large variance in body size means that large sample sizes are likely required to construct growth curves or to detect statistically significant differences in mean body mass between years.  8.3  Seasonal Pup Growth  I estimated growth rates from tagged and untagged pups weighed on four rookeries (Zapadni Reef, Polovina, Reef and Northeast Point). Procedures used to tag and weigh the pups are described in Chapters 7 and 9. Pups were tagged from 195766 when about 5-6 weeks old and weighed 2 weeks later. Additional samples of untagged pups were weighed during 1967-71, 1984 and 1987. A total of 17 116 pups have been weighed since 1957 from Aug 24 - Sept 5, of which 4 918 bore a tag or mark. The mass of pups weighed between Aug 24-Sept 3 was normally distributed and ranged from 1.0-18.8 kg with a mean of 8.7 kg (sd = 1.97, n = 17116). Two month old males are 13.5% larger than females (Table 8.2, Fig. 8.2) and appear to grow faster than female pups, given that the difference in size at birth was only 10% (Chapter 2). The different growth rates of the sexes is consistent with the marked sexual dimorphism of fur seals, and has been demonstrated for a related species, the Antarctic fur seal (Payne 1977, 1979; Doidge et al 1984a; Doidge and Croxall 1989; Croxall et al. 1987). Tagged pups presumably weighed less than untagged pups because those selected for tagging in mid August were smaller and younger than average pups, not because tagging affected growth (Chapter 7). A similar result was reported by Doidge and Croxall (1989) in their study of Antarctic fur seals. Since 1957, 90% of the pups were weighed from Aug 29 - Sept 3 (the only exceptions being 1984 and 1987). During this six day period, untagged male and female pups grew an average of 0.131 and 0.121 kg d  _ 1  , respectively (Fig. 8.3). Although  Chapter 8  Biased Estimates of Pup Mass  121  Figure 8.2: Relationship between the mass of male and female pups. Each data point indicates the mean mass of at least 30 pups weighed between Aug 29-Sep 3 and is specific to a given rookery in a given year (1957—71, 1984, 1987). Linear regressions were fit separately to data from tagged and untagged pups. Neither of the intercepts was significantly different from zero (tagged: <26 = 1.745, p = 0.092; untagged: t $ = 1.446, p = 0.154). Nor did the slopes of the two regressions differ from one another (i = 0.629, p = 0.266). Thus, the two data sets were pooled and fit with a zero constant (bottom right panel). The dashed line shows the expected 1:1 relationship. The significance of the regressions are contained within the panels. 5  8 3  Biased Estimates of Pup Mass  Chapter 8  122  Table 8.2: Mean mass of male and female pups weighed between Aug 29-Sep 3. The mean size of tagged pups, weighed from 1957-66, was compared to the mass of untagged pups weighed during the same time period using a Student's t-test. Probabilities associated with the sample statistic are enclosed in brackets (p). The mean mass of untagged males was compared to female mass over the years 1957-71, 1984 and 1987.  Mark  Sex  Mean Mass  no tag  M F  (kg) 9.47 8.30  tagged  M F  no tag tagged no tag tagged  Standard Deviation  i-test  df  (P)  1.98 1.74  35.10 (0.002)  12 196  8.43 7.49  1.84 1.60  19.06 (<0.001)  4 916  M M  9.26 8.43  1.98 1.84  16.36 (<0.001)  5 697  F F  8.14 7.49  1.75 1.60  14.362 (<0.001)  5 620  males appeared to grow faster than females, the difference was not statistically significant (slope: t n = 0.130, p = 0.448; elevation: t 0  ni  = 9.548, p < 0.001). Nev-  ertheless, the growth rates were high and indicate potential problems in trying to contrast the mean mass of pups collected in different years if they were not sampled on the same day. For example, an average sized male pup weighing 9.0 kg on Aug 29, would weigh 9.7 kg on the last day of sampling, Sept 3. Similarly, a female weighing 8.0 kg would gain 0.6 kg over the same 5 day sampling period. There is curiously little information available on the seasonal growth of northern fur seal pups. In 1952, Scheffer and Wilke (1953) weighed pups at birth and again on Oct 2. Assuming that 89 days had elapsed since the mean date of birth, the average growth rate was 0.096 kg d  _ 1  for males and 0.081 kg d  - 1  for females. Data  taken from Costa and Gentry (1986) shown in Fig. 8.1 (top panel) suggest that the growth over the first two months of life is linear and relatively slow (females grew  Biased Estimates of Pup Mass  Chapter 8  12 -  Males  North East Point Zapadni Reef Polvina Reef  10 -  &  CO CO  8  CO  ^  Q.  123  r=0.288 p-0.030  6  =J 12 CL  Females  c CO  S 10 8 -  6 -  r.0.312 P-0.018  i  1  29  30  August  r 31  T  1  1  2  T  3  September  Figure 8.3: Mean body mass of untagged pups weighed from Aug 29-Sep 3. Each point represents one rookery and one year (1957-71, 1984, 1987), where the sample size was greater than 30. Note that in pooling the data I am assuming that no bias is introduced by the type of scale used, and that the sample size was large enough to reduce the effects of feeding and fasting. Linear regressions were fit to the mean mass, rather than the raw values, to reduce the combined effects of annual differences in body sizes and sample sizes. The significance of the regressions is shown at the bottom of each panel.  0.051 kg d  - 1  and males 0.062 kg d ) . Note however that this growth curve is based _1  on a few pups that were repeatedly weighed prior to feeding which means that there is less information in the figure than might appear. Nevertheless, the regressions are still significant when the degrees of freedom are reduced to account for the repeated measurements. The growth rate from Aug 29 - Sept 3 (see Fig. 8.3 and above), is almost double that shown for July and August in Fig. 8.1 and may decline through the months  Chapter 8  Biased Estimates of Pup Mass  124  of September and October (Chapter 7). Growth rates of males weighed from Sept 2-Oct 3, 1962 averaged 0.092 kg d " and dropped to 0.046 kg d " from Oct 21  25 (Roppel et al. 1963, Chapter 7).  1  The suggestion from these three pieces of  information is that pup growth is nonlinear with mass increasing in a sigmoid fashion from birth to weaning. Several authors have concluded that fur seal growth from birth to weaning is linear. This has been reported for both Subantarctic (Kerley 1985) and Antarctic fur seals (Kerley 1985; Doidge et al. 1984a; Doidge and Croxall 1989). However, inspection of their data suggests growth is actually nonlinear and likely sigmoid (e.g. see Fig. 1 in Doidge et al. 1984a). 8.3.1  Sample Size Effects  Errors associated with sample sizes, behavioural segregation of different aged pups, and human error in randomly selecting pups for weighing are evaluated using the pup mass data base previously discussed and with independent studies. On all rookeries, except Northeast Point, there is a positive relationship between the mean mass and the number of pups weighed (Fig. 8.4). This correlation between body size and sample size suggests that pups were not randomly chosen for weighing. Pups gather together in small pods on the rookery and pile up during weighing. As suggested in Chapter 7 smaller pups may be taken from the tops of the piles. Evidence supporting this hypothesis was gathered in 1980 and reported by Roppel et al. (1981). They followed a shearing crew through four rookeries on St. Paul Island and weighed the pups selected for shearing from pods of 10-200 pups. When the shearers were finished, the remaining unsheared pups in the pod were weighed. Roppel et al. concluded that sheared pups generally weighed less than unsheared pups and noted the difference in weight was somewhat greater for males than for females. Their results suggest that smaller pups are easier to handle  Chapter 8  Biased Estimates of Pup Mass  125  r-0.526 p<0.001  Male Pups -a- Female Pups  CO  M).358 (teO.022  0  50  100  150  Sample Size Figure 8.4: Relationship between mean body size and the number of untagged pups weighed on three rookeries: Zapadni Reef, Polovina and Reef. A positive relationship between pup size and sample size was found for each rookery separately prior to pooling the data, except at Northeast Point.  and axe subconsciously chosen before larger ones. The correlation between body size and sample size shown in Fig. 8.4 could also be related to pups being segregated by size and age on the rookeries. Young and small pups may be in the periphery regions and easily captured for weighing. Attaining larger sample sizes may require going further into the rookery, thereby resulting in rounding up older and hence larger pups. Support for this theory comes from a 1980 study reported by Gentry and Francis (1981). They marked pups shortly after birth on two rookeries, and calculated the median ages of those marked pups captured many weeks later for shearing, as well as the median ages of those marked pups not captured. They found that captured pups were significantly younger (by 4-5 days) than non captured pups at both rookeries. Further inferences about the spatial distribution of pups can be made from studies of Antarctic fur seals, since comparable data from northern fur seals is lacking. In one study, Boyd (1989) counted the number of pups inland and on the beach. He found the rise in pup densities using the inland area was lagged behind  Chapter 8  Biased Estimates of Pup Mass  126  the beach density by about three weeks. It appears that pups dispersed from the high density beach to the low density inland areas. In so moving, they probably improved their chances of survival, given that mortality rates are greater at high density sites than at low density sites (Doidge et al. 1984b). Another mechanism that might produce spatial segregation relates to when pregnant females of different ages arrive on shore to give birth. It seems that Antarctic fur seals giving birth in the last quarter of pupping were shorter and lighter than average, and included a greater proportion of primiparae (Doidge et al. 1984b). In northern fur seals, primiparous females are also thought to return late and are known to give birth to small pups (Chapter 2). Perhaps there is a greater tendency for late arriving females to give birth in low density periphery areas as the main breeding beaches fill. Such a strategy could improve the chances of a young inexperienced female successfully raising her pup. Such an hypothesis is supported by the observation that higher rates of starvation of Antarctic fur seal pups were recorded on high density breeding beaches because disturbances induced by the fighting and boundary displays of breeding bulls disrupted the establishment of mother-pup bonds (Doidge et al. 1984b). The possibility of spatial segregation of different sized pups could confound attempts to compare the mass of northern fur seal pups sampled during the high densities of the 1950s with that recorded in the 1980s when the population was so low. It could also confound making comparisons between the size of Antarctic pups weighed at South Georgia as the population increases. Pups of different ages and sizes may be better mixed and less segregated at low densities. Furthermore, biologists are not restricted to the periphery regions of the rookery at low densities and could presumably round up older and larger pups down at the water's edge. Only the mass of northern fur seal pups from Northeast Point does not appear to be biased by the number of pups weighed. This might be explained by the  Biased Estimates of Pup Mass  Chapter 8  3 w  c  11 H  •  g  8  0)  7  O) CD CO  r-0.427 p-0.019  10  (0 CD  127  •  •  • Male Pups -a- Female Pups  •  • o•  l •  6 0  o  ° D  5000  10000  15000  Number of Tags Attached  Figure 8.5: Relationship between mean body size of a sample of tagged pups and the total number of pups bearing tagged. Each point represents one rookery in one year from 1957-66. Unlike the females, the male regression was not significant (r = 0.158, £28 = 0.624, p = 0.269). This might be explained by the large variance of male mass and the small number of means.  physical layout of this rookery which follows closely along the shoreline and has never extended far inland, such that a cross section of pups could always be weighed down to the water's edge. There is presumably a threshold level beyond which weighing more pups would not increase the mean mass. Just exactly what the sample size is and how it relates to rookeries of different densities is not clear and should be further investigated. Perhaps subsections of the data could be grouped and analyzed in the order collected. An additional piece of evidence supporting the segregation theory is the positive correlation between the mean weight of tagged pups and the number of pups previously captured and tagged from 1957-66 (Fig. 8.5). The data suggest that bigger pups were captured and tagged when large number of tags were attached.  This  probably occurred because pups had to be driven from deep within the rookery to obtain large numbers of pups.  Biased Estimates of Pup Mass  Chapter 8  8.3.2  128  Overview  Several factors should be considered when weighing fur seal pups. If for example, the intent is to make comparisons between the size of pups weighed in different years, then measurements should be made on the same day each year. This precaution assumes reproductive synchrony and controls for daily growth increments. It is also important to follow the same procedures when gathering and measuring pups, so that any hidden bias (i.e. if pups of different ages are not well mixed) is consistent between years.  Biases associated with spatial segregation of different aged pups  can be particularly troubling if the size of the population changes rapidly. At high densities only seals in the periphery regions of the rookery may be accessible, compared to a low density population where all individual can be obtained. Another important consideration is the sample size required to detect statistical differences in mean body mass, if they are indeed present. As a general rule, large samples are needed because of the large variability in body size that is attributed to such factors as the particular growth phase the pup is in, and whether or not a pup has fasted or recently fed. Since variability in body size increases with time, the later measurements are taken after birth, the greater the sample size required. Sexual dimorphism means that differences in females sizes can be detected with smaller sample sizes. Researchers may find that data from females is more reliable than male data, if large samples cannot be obtained. The sample sizes needed can be deduced from power analysis (e.g. Zar 1984), and should be estimated before investing time and effort into further pup weighing operations.  8.4  Summary  Many factors can bias estimates of pup mass and lead to incorrect conclusions, unless appropriate cautions are taken. Using data collected from tagged and untagged  Chapter 8  Biased Estimates of Pup Mass  129  northern fur seal pups, I assessed how milk consumption, the timing of birth, and the effects of growth and sample size influence the size of pups captured for weighing. Evidence is presented suggesting that pup mass may increase in a sigmoid fashion, with the most rapid rate of growth occurring at about age 2 months. This phenomenon can confound efforts to compare the masses of pups weighed on different days in different years, particularly if pups are weighed over the period of rapid growth. Variability in pup mass increases with time because growth rates of individuals vary and because the amount of milk pups consume increases with body size. Thus, sample sizes must be increased as the pups grow older, in order to detect statistically significant differences in mean body mass. There is also evidence that pups of different ages and sizes are not randomly distributed on the breeding beaches, and are not randomly selected for weighing. It appears that the first pups captured for weighing are younger and smaller than subsequent captures, possibly because smaller pups are easier to handle and are segregated to the peripheral rookery regions where sampling begins. These hidden biases, related to human error and fur seal biology, must be considered and controlled for when weighing pups.  Chapter 9  Food Abundance and Annual Fluctuations in the Growth of Pups 9.1  Introduction  Northern fur seal pups were weighed from 1957-71, when about 2 months of age (Roppel 1984). The purpose was to determine if body weight in autumn was related to survival and whether it influenced the size of future commercial harvests. No relationship was found (Roppel 1984). Today the time series of pup weights offers useful insight into changes in the size of suckling pups that occurred during a period of rapid population change. With the addition of data collected in 1984 and 1987, inferences can be made about the condition of the herd relative to food near the breeding Islands as the fur seal population declined from a peak production of 461 000 pups in 1955 to a low of 166 000 in 1983 (Lander 1980a, York and Kozloff 1987, Trites 1989). Studies of domestic mammals have shown that maternal food intake alters the quality and quantity of milk produced, which in turn affects the offspring's growth (Stern et al. 1978; De Lange et al. 1980; Hodgson et al. 1980; King et al. 1980; Stockdale et al. 1981; Holdroyd et al. 1983). Annual differences in the size of fur  130  Food Abundance & Pup Growth  Chapter 9  131  seal pups might therefore be influenced by maternal competition for food, and thus act as an index of population status relative to carrying capacity (Chapman 1961; Eberhardt and Siniff 1977). Support for these views comes from several fur seal studies that have compared annual fluctuations in pup growth with environmental conditions  (e.g. New Zealand fur seals, Arctocephalus forsteri: MattHn 1981; Gala-  pagos fur seals, seals,  A. galapagoensis: Trillmich and Limberger 1985; and Antarctic fur  A. gazella: Doidge et al. 1984a; Croxall et al. 1987; Doidge and Croxall 1989).  Lactating northern fur seals follow a feeding-nursing rhythm over 3-4 months that typically involves spending an average of 5 days foraging for food followed by 2 days ashore suckling their pups (Gentry and Holt 1986). Most females swim 100400 km west and south of the Pribilof Islands to feed primarily on walleye pollock  (Theragra chalcogramma) along the continental slope (Antonelis et al. 1986; Gentry et al. 1986; Hacker and Antonelis 1986; Perez 1986; Perez and Bigg 1986; Loughlin and Bengtson 1988; French et al. 1989; Livingston 1989; Lowry et al. 1989). The amount of milk the mother produces and transfers to her single pup is likely a reflection of the quantity and quality of fish and squid she captures while feeding. My goal is to determine what changes occurred in the size of nursing pups as the size of the population declined over three decades, and to relate these changes to the abundance of pollock in the Bering Sea. I begin by outlining the methods used since the late 1950s to capture and weigh northern fur seal pups on the Pribilof Islands. I then try to control for possible errors, associated with varying sample sizes and weighing pups on different days, before comparing the sizes of pups collected from various rookeries in different years. Finally, I consider changes that have occurred in the fur seal food base and in the breeding success of seabirds in light of the fur seal findings.  Chapter 9  Food Abundance & Pup Growth  9.2  Materials and Methods  9.2.1  Weighing  132  On average, pups were about 7-8 weeks old when weighed in late August or early September. The earliest weighing occurred in 1987 from August 24-31; the latest in 1984 from September 3-5. In all other years (1957-71) pups were weighed from Aug 29-Sep 3. Some of the pups weighed from 1957-66 bore tags that had been attached when the pups were 5-6 weeks old. Data from tagged pups are analyzed separately from untagged pups because tagged pups are smaller (Roppel 1984) and are thought to be younger than untagged pups (Chapter 7). Pups were weighed annually on four rookeries (Zapadni Reef, Polovina, Reef and Northeast Point) from 1957-71, and in 1984 and 1987. The only year when all four rookeries were not sampled was 1958 when the scales were found to be out of order after work at Northeast Point had been completed (Abegglen et al. 1958). Details of the sampling procedures are contained in Chapters 7 and 8. A total of 17 116 pups were weighed from Aug 24 - Sep 5 since 1957, of which 4 918 bore a tag or mark. Changes in the type of scale and method employed to weigh pups from 1957-66 were described by Roppel et al. (1966) as follows. 'In 1957 and 1958, a metal cone for holding the pups was attached to a dial spring scale suspended from a wooden tripod. The spring scale, however, was too sensitive to movement of the pups, and the cone caused the pups to move considerably in their attempts to escape. Both problems were partially solved in 1959 by putting the pups in burlap bags sewed to construction-steel hoops and placing the bag containing the pup on a platform scale. Though the platform scale had no dampeners, it was less sensitive than the spring scale to movement of the pups, and the pups were less inclined to struggle when confined in the bag. The bags, however, changed in weight as they dried or  Chapter 9  Food Abundance & Pup Growth  133  absorbed rain water. Frequent weighing of the bags and corresponding adjustments of the scale eliminated much of the error from this source. In 1963 and 1964, the same platform scale was used, but the pups were placed in 20-gallon plastic garbage cans. The pups tended to move more when in this container than when confined in the burlap bags. The method of holding the pups, therefore, was changed in 1965 and a new platform scale, with dampeners, was used. One man stood on the scale and held each pup during weighing, a method that effectively eliminated movement of the pups. The weight of the man was recorded after each series of 25 weighings, for later subtraction."  1  This method was used until 1971. In 1984 and 1987, pups  were placed in nets and weighed from a small spring scale hanging from a tripod. Depending upon the year of weighing, pup mass was rounded to the nearest 0.10 kg (1957, 1963, 1965), 0.20 kg (1959-62, 1964), 0.25 kg (1987) or 0.50 kg (1958, 1966-71, 1984). 9.2.2  A n n u a l Variations in Weight  Comparisons of annual variations in mean pup mass were restricted to the period Aug 29-Sep 3, to reduce the effect of daily growth increments. This is valid since peak pupping varies little from year to year (Chapter 4). Trends in the mean mass of pups nursing on the four rookeries from one year to the next were estimated by a locally weighted smoothing routine (lowess, see Cleveland 1979). The data were grouped into overlapping two day blocks to further control for size differences related to seasonal, rather than annual growth changes. Sample sizes less than 30 were excluded to reduce the effect of over- or under-sampling fasted or recently fed pups. As previously mentioned, the weights of pups bearing tags were analyzed separately from those of untagged animals because the two groups of pups are known to differ. 'from Roppel et al. 1966  Chapter 9  9.3  Food Abundance & Pup Growth  134  Results and Discussion  Mean size of pups varied considerably from one year to the next (Figs. 9.1 and 9.2). Even within years, the size difference between rookeries was sometimes huge (e.g. 1964, bottom right panel of Fig. 9.1). No one rookery had consistently light or heavy pups, although there is some suggestion, based on ranking the annual mean mass on each rookery from lightest to heaviest, that untagged pups on Reef and Northeast Point may, on average, have been slightly heavier than pups from the other two rookeries. In theory, the mean size of pups born on all four rookeries should be the same, given that their mothers prey upon a common food base and that most pups are born on the same date in all rookeries. Possibly there are genetic differences between the different rookeries because fur seals have a strong tendency to home and reproduce on the rookery of birth. But a more likely explanation for the size differences lies with small sample sizes in conjunction with other sampling biases (Chapter 8). Comparing the sizes of pups weighed over the same two day period from 195771 revealed a consistent trend in pup size as the date of sampling moved from Aug 29-Sep 3 (Figs. 9.1 and 9.2). Unfortunately conclusive statements about long term changes in mean body size cannot be made when only two days of sampling are considered because of the small numbers of samples and the huge variability in mean body size. These figures illustrate the difficulty of assessing annual changes in pup mass when few samples are available, particularly when data from any one year is missing. Pooling the mean mass of pups weighed on all four rookeries from Aug 29 - Sep 3 shows a gradual increase in pup size from 1957 to 1971 (bottom right panels of Figs. 9.1 and 9.2). The fluctuations in pup mass may reflect periodic changes in the food base and oceanographie conditions in the Bering Sea (see discussion to follow).  Chapter 9  Food Abundance & Pup Growth  Figure 9.1: Mean mass of untagged pups weighed over a two day period from 1957-71. The bars indicate 95% confidence intervals for the mean mass of male and female pups sampled at each rookery. The data were jittered to reduce overlap, and were smoothed with a locally weighted curve (lowess / = 0.67). Data are pooled in the bottom right panel and are shown without confidence limits to reduce clutter (lowess / = 0.40).  135  Food Abundance & Pup Growth  Chapter 9  • Northeast Point A Polovina  10-  136  O Reef • Zapadni Reef  Aug 29-30  Aug 30-31  Aug 31-Sep 1  Sep 1-2  9 8 -  CD to Q. D  CO CO CO  c CO CD  7 109 8 7 10-  Sep 2-3  Aug 29-Sep 3  8 -  1 —i 1958  1  1961  1  1  1  1964  1967  1970  Year  '  1  1  1958  1961  1  1  1964  1967  Year  r 1970  Figure 9.2: Mean mass of tagged pups weighed over a two day period from 1957-66. The bars indicate 95% confidence intervals for the mean mass of male and female pups sampled at each rookery. The data were jittered to reduce overlap, and were smoothed with a locally weighted curve (lowess / = 0.67). Data are pooled in the bottom right panel but are shown without confidence limits to reduce clutter (lowess / = 0.40).  Chapter 9  Food Abundance & Pup Growth  137  This of course assumes that pup growth is directly related to the amount of energy received from a pup's mother which in turn is directly related to the amount and quality of prey near the Pribilof Islands. Other factors besides maternal milk that can affect pup size are the prevailing weather and rookery crowding. Pup growth could be impeded at high population densities because of diseases, trampling and biting. Similarly under inclement weather conditions, pups might allocate a greater portion of their energy reserves to maintenance rather than to growth. Unfortunately there is little information on the effects of rookery crowding on growth. As for weather, conditions recorded on St. Paul Island since 1956 were not extreme enough to have significantly affected healthy, average sized pups (Chapters 5 and 6). Thus it would seem that the quantity and quality of milk is probably the critical component of growth. The major prey of lactating females appears to be young walleye pollock ranging in length from 4 to 40 cm (Perez 1986; Perez and Bigg 1986; Hacker and Antonelis 1986; Livingston 1989). Population estimates by age group for pollock in the eastern Bering Sea based on cohort analysis (1964-75) indicate a low biomass in 1964—65 which increased to peak numbers in 1968-69 and declined again in 1972-73 (Bakkala et al. 1987). Unfortunately estimates of stock size do not extend back to the start of pup weighing in 1957. The changes in pollock abundance (Fig. 9.3) appear to mimic the increase and decrease in pup mass shown in Figs. 9.1 and 9.2 (bottom right panels). The most likely cause of variation in pollock year class strength and population abundance appears to be environmental conditions, although the exact mechanism is not understood (Laevastu 1983; Bailey et al. 1986; Bakkala et al. 1987; Bulatov 1989; Chapter 11). Other species in the Bering Sea also seem to show comparable fluctuations in breeding success. This has been documented by Springer and Byrd (1988) for some populations of sea birds that feed on age 1-year pollock. Data collected since 1975  Food Abundance & Pup Growth  Pol lock  Chapter 9  138  16 12 -  Age  CO  Billions  o  840-  ^  1964  1  1 968  i 1 972  1  1976  1  1 980  T  1 984  H 1988  Year Figure 9.3: Estimated number of 3-year old walleye pollock in the eastern Bering Sea in billions of fish (data are from Wespestad and Traynor 1990).  suggest the possibility of a 7-8 year cycle in the reproductive success of black legged Mttiwakes  (Rissa tridactyla) on the Pribilof Islands that is correlated with sea surface  temperature. The increasing size of pups from 1957-71 occurred as the numbers of pups born on St. Paul Island declined from 420 000 in 1957 to 250 000 in 1971 (Lander 1980a; Trites 1989). A major portion of the population decline during this period was due to the harvesting of 315 000 mature females between 1956 and 1968, in addition to a series of years of poor juvenile survival (York and Hartley 1981; Eberhardt 1981; Trites and Larkin 1989). It thus appears that the size of pups may have increased because the smaller herd size reduced competition between lactating females foraging for food near the Pribilofs. Tagged pups weighed less than untagged pups and underwent more pronounced fluctuations in body mass (Figs. 9.1 and 9.2). If pups captured for tagging were born later in the breeding season there may have been less food available to their mothers, or food may not have been so easily obtained. This could account for the stronger fluctuations in tagged body mass.  Chapter 9  Food Abundance & Pup Growth  139  It is difficult to compare the estimated size of pups recently weighed to the historic time series, because pups were weighed much later than normal in 1984 (Sep 3-5) and much earlier than normal in 1987 (Aug 24-31). Furthermore, the size of each rookery is now so small that previous biases associated with the segregation of different sized pups on the rookery and the number of pups captured are probably no longer present (see Chapter 8). Pups of different ages and sizes do not seem to be randomly distributed on high density rookeries and were probably not randomly selected for weighing during the 1950s and 60s. It appears that the first pups captured for weighing are younger and smaller than subsequent captures, possible because smaller pups are segregated to the periphery regions where sampling begins (see Chapter 7 and 8). Nevertheless, I believe that valid comparisons can be made between the mass of pups weighed in different years at Northeast Point because this rookery always followed the water's edge and never extended far inland, unlike the other rookeries. Data from Northeast Point indicate that pups are probably no bigger now than they were in the late 1960s (Fig. 9.4). On Zapadni Reef however, it might be erroneously concluded from Fig. 9.4 that pups are much bigger now than ever in the past, because the rookery no longer extends as far inland as it did when the first weights were recorded in the late 1950s. From the data I analyzed, it appears that northern fur seal pups are as heavy now as they ever were, and are not limited by food. This view is supported by data showing a decrease in the length of feeding trips by lactating females since the 1960s (Gentry et al. 1977; Gentry and Holt 1986; Loughlin et al. 1987). It could thus be incorrectly argued that fish stocks used by lactating fur seals near the Pribilofs are abundant. Instead the findings only imply that the adjacent food is sufficient to support the currently low and declining Pribilof fur seal population. Given the variable recruitment of pollock and the large commercial pollock fishery in the Bering Sea, it is not known whether the number of pups born during the  Chapter 9  Food Abundance & Pup Growth  140  Figure 9.4: Mean mass of tagged and untagged pups weighed on four rookeries from 1957-71 (ZRF - Zapadni Reef, NEP - Northeast Point, POL - Polovina, REF - Reef). Each data point represents the mean of mass of pups (males and females combined) in one year. Confidence limits for the means are shown in Figs. 9.1 and 9.2. Each data set was smoothed with a locally weighted curve (lowess / = 0.5).  1940s and 50s could be sustained today. Nor is there any evidence that fur seals have ever been abundant enough to affect their food density. But studies of Pribilof sea birds that depend almost entirely upon age 1-year pollock during the breeding season suggest the ability of the food base to support a large population of sea birds appears to have been reduced since 1975 and is probably much lower than during the 1960s (Springer and Byrd 1988). I believe pup weights are a good index of fur seal population condition, but I caution that they shed light only on the status of the population relative to carrying capacity during the breeding season. Changes in pup mass undoubtedly reflect the amount of food near the breeding beaches and the extent of female competition while feeding. However, pup mass seems to have little affect on subsequent survival  Chapter 9  Food Abundance & Pup Growth  141  given that pup survival at sea (Trites 1989) did not change in parallel with changes in pup mass on land.  9.4  Summary  Northern fur seal pups were weighed at various intervals between 1957 and 1987 when approximately two months old. The mean mass of the pups born on St. Paul Island increased from 1957-71 as adult females were harvested (for the first time in recent history) and the population declined. The data also suggest fluctuations in pup mass that may relate to fluctuating pollock abundance in the Bering Sea. Pups weighed in the 1980s were as heavy as those weighed in the late 1960s. This relationship suggests that the food supply near the Pribilofs is sufficient to support the currently low and declining fur seal population. Pup mass is likely a good index of food abundance during the breeding season, and of a population's breeding size relative to the maximum level that can be supported by the resources near the breeding beaches.  Chapter 10  Seasonal Growth Fluctuations During the Fur Seal Migration 10.1  Introduction  From 1958-74, Canada and the United States, as members of the North Pacific Fur Seal Commission, recorded morphometric (body) measurements from over 18 000 northern fur seals shot at sea between California and the Bering Sea. The extensive data collection contains valuable information about the growth of northern fur seals and offers intriguing insights into the general processes underlying pinniped growth. Quantitative descriptions of fur seal growth from the pelagic data base can be used to explore functional and evolutionary links among demography, energetics, and sociobiology. Pinniped growth studies can also be used to determine population biomass or to identify changes that might have occurred in the marine ecosystem (Bryden 1972; Payne 1979; McLaren 1981; McLaren and Smith 1985). Scheffer and Wilke (1953) were the first to describe growth of male and female northern fur seals from animals killed on land. Later, Taylor et al. (1955) and Nagasaki (1961) constructed growth curves for females collected at sea, and noted that pregnant females were longer than non-pregnant females of the same age. This phenomenon was confirmed in subsequent pelagic studies by Fiscus et al. (1964, 1965)  142  Chapter 10  Growth Fluctuations & Migration  143  and further investigated by Bigg (1979) who also considered seasonal changes in the size of pregnant females. This was followed by Lander (1979a) who summarized the 1958-74 pelagic data by month and constructed monotonic growth curves by age and sex. Growth curves have also been drawn for Soviet populations of northern fur seals (Ito 1969; Bychkov 1971a; Chelnokov and Chugunkov 1971; Muzhchinkin 1976). In the current study, I describe seasonal fluctuations in the growth of northern fur seals breeding on the Pribilof Islands. Daily growth patterns and seasonal fluctuations in the size of immature males and females have not been described in detail before, and axe done so here. I also consider whether fur seal growth during the annual migration is dependent on geographic feeding location. Again this has not been previously considered. I begin by briefly reviewing fur seal biology and outlining the major events in their life cycle. I then describe the methods used to collect and analyze the data. Consideration is given to the factors that give rise to seasonal fluctuations in the animals' length and mass. This is followed by a discussion on how growth rates vary between seals of different ages and different reproductive statuses. Finally, I consider growth processes in other pinniped species, and examine the relationships between growth, feeding location and timing of migration.  10.2  Methods  10.2.1  F u r Seal Biology  All of the northern fur seals collected at sea from 1958-74 were believed to have come from the Pribilof Islands, Alaska, although a few stray animals could have originated from Soviet populations (Lander 1980b). The annual fall migration of fur seals extends southward from the Bering Sea to California. The animals spend about two-thirds of their life at sea and the remaining third on or near the Pribilofs  Chapter 10  Growth Fluctuations & Migration  (Lander and Kajimura 1982).  144  It appears that the timing of migration and the  distance travelled is a function of age and sex (Bigg 1986, 1990). In general, females go farther south than males, and older animals farther south than younger ones. The first to return north are the mature bulls in late May and early June (Kenyon and Wilke 1953; Fiscus 1978). They axe followed by pregnant females in late June and early July (Bartholomew and Hoel 1953; Peterson 1968). Over 75% of the females give birth during a three week period from the end of June to mid July (Chapter 4). The summer return schedule appears to be a function of age and maturity, with younger animals arriving on the Pribilofs progressively later in the breeding season (Bigg 1986). Mature bulls fast for 1-2 months while defending breeding territories (Kenyon and Wilke 1953; Fiscus 1978). Immature males and some immature females gather together on haulouts adjacent to the breeding beaches. During the years of pelagic sampling, young males were regularly rounded up from the haulouts for harvesting and there was a substantial land harvest of 315 000 females from 1956-68 (Lander 1980a). The animals begin moulting in August (Scheffer and Johnson 1963; Bychkov 1971b) and begin their southward migration in late October and November as the rookery structure breaks down and pups are weaned (Peterson 1968). 10.2.2  Collection of D a t a  Methods used to collect and prepare the samples are described by Lander (1980b) and are briefly summarized as follows. Fur seals were shot from government and chartered vessels used to survey waters in the animal's feeding range from California to the Bering Sea (Fig. 10.1). The number of fur seals collected from 1958-63 was subject to quotas whereby a minimum number were to be collected each year. The times and areas of collections were not specified. Thus research during the quota years was concentrated in areas of known fur seal abundance, based on the experience of commercial pelagic sealers before 1911 and pelagic research expeditions in 1952  Growth Fluctuations & Migration  Chapter 10  145  - 61" N C A N A D A  • 51°  41°  • 31° 175° W  165°  155°  145°  135°  125°  115°  Figure 10.1: The seven regions where fur seals were collected from 1958-74 (adapted from Kajimura, 1985). and 1955. After the quota was removed in 1964, research vessels tended to follow systematic transects in some areas, but were still somewhat influenced by prior knowledge of expected fur seal distributions. Seals were sought during daylight hours and shot. They were processed as soon as possible after being taken aboard, before rigor set in. The seals were dropped onto a 'cradle' measuring board with their backs down. Length was measured from the tip of the nose (positioned to touch zero) to the tip of the tail.  Care was  taken not to unduly stretch the animal. After 1967, the total distance from zero to the tip of the tail was noted on U.S. vessels, and the distance between the tip of the snout and zero was then subtracted from this total to give the length of the  Chapter 10  Growth Fluctuations & Migration  146  animal. The seals were weighed with spring, torsion, or platform scales (usually to the nearest 0.5 kg). Jaws were taken and sent to laboratories to remove the canine teeth and age the animals. Reproductive condition (either nulliparous, primiparous, or multiparous) of the females was determined from field examination of the uterine horns and superficial examination of ovaries for ruptured follicles, until 1962 (by the U.S.). Aboard Canadian vessels (1958-74), and U.S. vessels (after 1962), the entire female genital tract was removed, placed on a metal stretcher to harden and preserved in 10% formalin for later examination in the laboratory. 10.2.3  Analysis of D a t a  The method used by Canada and the United States to assign ages to the pelagic samples was not appropriate for analysing the growth data. The standard practice was to assign a birthday on January 1, ignoring the true biological age of the animal (Lander 1980a). In actual fact, birthdays were assigned in November, such that a pup born in July and killed only 5 months later was recorded as 1 year old. Similarly, animals killed at 17 and 28 months were both considered to be 2 years old. I therefore corrected the data to reflect the true biological ages of the samples, then determined the ages of the animals in days using July 1 as the mean date of birth. I considered three classes of animals, consisting of 2 008 males, 6 493 non-pregnant females, and 9630 pregnant females. The pregnant females included recently postpartum animals taken mainly during July and August in the eastern Bering Sea. Data from all years and areas of the eastern Bering Sea and eastern North Pacific were combined. They were plotted (i.e. length versus mass) to identify outliers which were later verified with the original field notes. Growth in body length and mass over time was described by robust locally weighted regressions (lowess, Cleveland 1979). The lowess algorithm requires choos-  Chapter 10  Growth Fluctuations & Migration  147  ing a smoothness parameter / which is a number between 0 and 1. As / increases, the fitted curve becomes smoother. I began by choosing a small / value, and increased it till I achieved a curve that was as smooth as possible, yet did not distort underlying patterns visible by eye in the data. I then plotted and smoothed the residuals, to ensure no residual structure remained. The relationship between body length (i) and mass (M) was estimated by linearizing the relationship M = aL  b  such that log(M) = log(a) + b log(i) where b is  the slope and log (a) is the intercept. It has been suggested that the geometric mean of the regression of mass on length and the inverse of the regression of length on mass should be used to give the appropriate linear regression for mass-length comparisons because length is not truly independent of mass (Kicker 1979). However, others have indicated that ordinary least squares regression is appropriate and easier to interpret than geometric mean regression (Sprent and Dolby 1980; Cone 1989). Thus I conducted ordinary least squares regressions. I analyzed seasonal fluctuations in body length and weight for four classes of animals: immature males (1.5 - 4.5 y), immature females (1.5 - 4.5 y), pregnant females (4.5+ y) and non-pregnant females (4.5 y). Samples of mature males and +  yearlings were insufficient to detect seasonal changes in body size. I considered only the period January to September, because fewer than 10 seals were collected for each month from October to December. Each time series was again smoothed using the robust locally-weighted lowess algorithm. To determine whether fur seal growth was independent of geographic feeding location, I plotted the area sampled (grouped according to Fig. 10.1) against the Julian date when sampled. The data were smoothed (lowess) to show when and where the previously mentioned four classes of fur seals migrate. These migration plots were then compared with the seasonal growth curves to determine the depen-  Chapter 10  Growth Fluctuations & Migration  148  dependence of growth on feeding location.  10.3  Results  10.3.1  Seasonal Fluctuations  Growth curves for males, non-pregnant females, and pregnant females are shown in Figs. 10.2 to 10.4. Immature males and females show seasonal changes in body weight that peak at progressively earlier times in the year as the animals grow older. The data for mature males axe scant and little can be concluded from them. The indication for mature females is that non-pregnant females do not experience sharp fluctuations in body weight. Weight change in pregnant females reflects the development of the fetus, peaking near the mean date of birth, July 1. An unexpected growth characteristic evident from the data isfluctuationsin body length. Within each age, length not only increases seasonally, but also decreases (top panels of Figs. 10.2 - 10.4). As with weight, there is a tendency for the decrease to be more seasonally protracted with increasing age. The seasonal changes in body size are seen most clearly in the plots of seasonal growth (Figs. 10.5 and 10.6). Immature males grow from the beginning of May until the first week of July, after which they too lose body mass and length. Few males were captured from January to March, but it appears that they lose weight throughout the spring months. This is certainly the case for immature females. The female seasonal growth curve appears to be lagged by about three weeks behind that of the males (i.e. immature females grow from the end of May until late July; see Fig. 10.5). Mature nonpregnant females (4.5 y) lose weight from January to mid April. +  This gradual weight loss is rapidly regained in 30 days from mid April to mid May. The large fluctuations in the size of pregnant females (Fig. 10.6) are related to the growth of the developing fetus. The most rapid gain in weight begins in early April  Growth Fluctuations & Migration  Chapter 10  149  150 O)  to 100 to CO  50 0 -  i 6  i 9  Age (y)  r 12  15  T 18 50  100  150  T 200  Length (cm)  Figure 10.2: Length and mass of 2 008 males by age in days. A small amount of random noise was added to each variable to reduce the overlap caused by measurement roundoff. The vertical reference lines mark the mean birth date, July 1. The growth curves were extracted by lowess (/ = 0.15). Note the seasonal changes in length and weight of young animals that peak at progressively earlier times in the year as the animal grows older. Note also the absence of data from seals aged 1-1.5 y and the rapid increase in body mass that occurs between the ages of 4 and 5 y.  Chapter 10  Growth Fluctuations & Migration  150  Figure 10.3: Length and mass of 6493 non-pregnant females by age in days. A small amount of random noise was added to each variable to reduce the overlap caused by measurement roundoff. The vertical reference lines mark the mean birth date, July 1, and the growth curves were extracted by lowess (/ = 0.08 left panels, / = 0.04 right panel). Note the seasonal changes in length and weight of young animals that peak at progressively earlier times in the year as the animal grows older, until disappearing when sexually mature. Note also the absence of data from seals aged 1-1.5 y.  Chapter 10  Growth Fluctuations & Migration  151  Figure 10.4: Length and mass of 9 630 pregnant and postpartum females by age in days. A small amount of random noise was added to each variable to reduce the overlap caused by measurement roundoff. The vertical reference lines mark the mean birth date, July 1, and the growth curves were extracted by bwess (/ = 0.07 left panels, / = 0.04 right panel). Note the seasonal changes in length and weight (which includes the fetus if present), peaks near July 1.  Chapter 10  Growth Fluctuations & Migration  152  Figure 10.5: Seasonal changes in the growth of immature males and females. Length and mass data were pooled for all animals between the ages of 1.5 and 4.5 y, and plotted against the day sampled. The smoothed lowess curves (male / = 0.33; female / = 0.25) show a decrease in length and mass of females through the spring, followed by an increase in the growth of both sexes beginning in May and peaking in July.  Chapter 10  Growth Fluctuations & Migration  153  Figure 10.6: Seasonal changes in the size of pregnant and non-pregnant females. Length and mass data were pooled for all animals older than 4.5 y, and plotted against the day sampled. The smoothed lowess curves (pregnant / = 0.20; non-pregnant / = 0.30) show a small seasonal change in the size of non-pregnant females and marked changes in the size of pregnant females associated with fetal growth and parturition.  Chapter 10  Growth Fluctuations & Migration  154  and peaks in mid June. Pups are born from late June to late July. Thus the growth peak occurs 1-2 weeks earlier than parturition time. This likely results from the fact that pregnant females swim directly to land, where upon they give birth and thus lose weight. Post-partum females stay with their pup for the first week and will later spend an average of 5 days at sea feeding and 2 days ashore suckling their pups (Gentry and Holt 1986). Thus they would be more vulnerable to capture and hence would be over-represented in the samples. Male and female growth curves are compared in Figs. 10.7 and 10.8, and estimates of means body size given in Table 10.1. The first figure highlights the strong seasonal component of growth, while the second represents the traditional monotonic growth paradigm. In both cases, it is clear that pregnant females are larger and longer than non-pregnant females. Curiously, size in old pregnant females declines with age in contrast to the increasing size of non-pregnant females. This is most visible in Fig. 10.8, and is related to fact that a senescent decline occurs in the size of young carried by females beyond the age of 11 y, despite the fact that the mother continues to grow (Chapter 2). 10.3.2  A l l o m e t r i c Relationships  The relationship between body mass (M in kg) and length (X in cm) is described by the equations: M = 4.318 x 10" X 5  2  825  for males, M = 6.081 x 10" Z 5  pregnant females and M = 9.794X 10~ Z 5  2  6 6 6  2  7 4 0  for non-  for pregnant females. Males and non-  pregnant females appear to follow the same allometric relationship until males reach puberty between the ages of 4 and 5 y (Fig. 10.8). After this age the male's allometric relationship departs significantly from that of the females. There is a remarkable consistency between postnatal and prenatal allometric curves. In the case of nonpregnant females, the equation describing the relationship between body mass and length is virtually identical to that for female fetuses (M — 6.607 X 1 0 Z -5  2  740  ,  Chapter 10  Growth Fluctuations & Migration  155  — Males — Pregnant Females — Non-Pregnant Females  Age (y)  Length (cm)  Figure 10.7: Smoothed growth curves generated with bwess using large / values that remove seasonal growth fluctuations (/ = 0.35 left panels, / = 0.15 right panel). The vertical reference lines mark the mean birthday, July 1. Note that growth appears to be indeterminate and that pregnant females are bigger than nonpregnant females. Also note that the length-weight relationship is identical for males and non-pregnant females until the male growth spurt begins between the ages of 4-5 y.  Chapter 2). 10.3.3  G r o w t h and M i g r a t i o n  The migration of fur seals is displayed in Figs. 10.9 and 10.10 and confirms the summaries made by others. Females migrate further south than males, with pregnant females going further south than non-pregnant females. The northward migration begins in February. The return of pregnant females to the Pribilof Islands occurs in a highly regular fashion. Non-pregnant females appear to be briefly delayed in  Chapter 10  Growth Fluctuations & Migration  156  — Males — Pregnant Females — Non-Pregnant Females  Age (y)  Length (cm)  Figure 10.8: Comparison of male, pregnant female and non-pregnant female growth curves shown in Figs. 10.2-10.4. The vertical reference lines mark the mean birth date, July 1. Note the timing of the seasonal peaks in length and mass.  Oregon and Washington and complete their migration a few weeks after the pregnant females. Immature females delay their return by a month by remaining off the coast of Washington. Young males also congregate off Washington, but depart for the Pribilofs before the immature females. The Washington coastal waters are a major transitional zone in the migration of northern fur seals as shown in Fig. 10.10. Comparing this plot to the seasonal growth patterns (Fig. 10.11) indicates that fur seals maintain or lose weight while in Washington waters and areas further south. Only after each age group of northern fur seals has left Washington, do they begin to grow and recover their lost body mass. In all cases, the seasonal growth peak occurs when the animals arrive on the  Chapter 10  Growth Fluctuations & Migration  157  Figure 10.9: Date and location where fur seals of different ages and reproductive statuses were collected. A small amount of random noise was added to the data when plotted to show the relative density of data and reduce the overlap of data points. The location numbers are those used to code the original data (as shown in Fig. 16.2) and are arranged from south (ll:Ca!ifornia) to north (88:Bering Sea). The smoothed curves (lowess, / = 0.33) indicate the timing of the northward migration to the Pribilof Islands. Note that the two age categories were 1.5-4.5 y (2-4) and > 4.5 y (5 ). +  Growth Fluctuations & Migration  Chapter 10  158  Table 10.1: Mean length (cm) and mass (kg) of males, non-pregnant females and pregnant females (including the fetus). The data are from the smoothed growth curves shown in Fig. 10.7 which removed the seasonal growth component.  Age 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  males 87.2 100.5 110.9 119.9 132.5 143.2 153.1 162.0 169.6 176.4 183.0 188.7 194.9 201.1 206.7 212.9  -  Length non-pregnant 79.4 93.1 103.1 110.6 114.8 118.2 120.1 121.7 122.8 124.4 125.5 126.0 126.3 126.8 127.4 127.8 127.9 128.2 128.5 128.8 129.0 129.1 129.2 129.4 129.5  pregnant  -  115.9 117.4 120.2 122.5 124.0 124.9 125.7 126.7 126.6 127.3 128.3 128.2 128.2 129.0 129.4 129.3 129.5 129.7 130.0 130.3  -  -  males 13.1 19.1 25.5 31.3 43.2 58.3 72.8 87.3 100.1 113.1 125.6 139.7 151.8 167.2 179.1 192.9  -  Mass non-pregnant 9.8 14.8 19.2 23.8 26.8 28.7 30.7 32.0 33.2 34.5 35.6 36.6 37.6 38.4 39.0 39.5 40.0 40.4 40.6 40.9 41.1 41.2 41.4 41.6 41.7  pregnant  -  27.4 30.7 33.6 35.9 37.9 38.4 39.5 40.8 40.8 40.9 42.0 42.7 42.6 43.0 43.4 43.5 43.5 43.6 43.7 43.8  -  Pribilof Islands.  10.4  Discussion  10.4.1  Pinniped Growth  Pinniped growth is commonly considered to be a monotonically increasing function of age (e.g. Scheffer and Wilke 1953; Boulva and McLaren 1979; Lines et al. 1981; Lander 1981; McLaren and Smith 1985). This view is clearly an oversimplification  Growth Fluctuations & Migration  Chapter 10  159  Date Figure 10.10: Comparison of the timing of migration through the North Pacific for different components of the fur seal herd shown in Fig. 10.9.  which appears to have been perpetuated by a scarcity of data or an absence of samples outside the breeding season. In some cases, researchers have masked seasonal growth changes by tabulating body size by year. Seasonal fluctuations in body mass have been noted for many species of phocids and otariids in connection with reproduction in mature animals. Pronounced changes in the seasonal fat content of sexually mature phocids have been noted for grey seals  Halichoerus grypus (Nordoy and Blix 1988) harbour seals Phoca vit-  ulina (Boulva and McLaren 1979; Ashwell-Erickson and Eisner 1983; Pitcher 1986), hooded seals  Cystophora cristata (Bowen et al. 1987), ringed seals P. hispida (Ryg  et al. 1990), southern elephant seals Mirounga leonina (Bryden 1968), northern elephant seals  M. angustirostris (Ortiz et al. 1984; Costa et al. 1986), and harp seals  Pagophilus groenlandicus (Sivertsen 1941; Sergeant 1973b; Innes et al. 1978; Stewart and Lavigne 1984). In all cases, it seems that the change in phocid body mass is associated with fat and not body core during lactation, rutting and moulting, when the animals eat little or nothing.  Growth Fluctuations & Migration  Chapter 10  130  160  -  — — — —  Pregnant Females Non-Pregnant Females Immature Males Immature Females  Figure 10.11: Comparison of seasonal changes in the size of immature males, immature females, pregnant females and non-pregnant females shown in Figs. 10.5 and 10.6.  With the exception of hooded seals, none of the phocid studies report seasonal fluctuations in the size of immature animals. Ryg et al. (1990) suspect that immature hooded seals gain more core mass than blubber in the spring and early summer. They suggest that seasonal changes in body mass are regulated by hormone levels associated with reproduction, and are hence more pronounced in adult seals than in immature seals. It is possible that seasonal changes in the size of other immature phocid species occur but may not have been observed because of incomplete sampling. Further insight might be gained from studies of captive animals. For example, early indications for immature harp seals fed in captivity suggest that the animals control their food intake and experience seasonalfluctuationsin body size (D.M. Lavigne, pers. comm.). As with phocids, most of the otariid studies have also tended to focus upon  Chapter 10  Growth Fluctuations & Migration  161  sexually mature individuals, particularly mature males. They have shown seasonal changes in the size of Australian sea Hons southern sea Hons  Neophoca cinerea (Walker and Ling 1981),  Otaria byronia (Vaz Ferreira 1959), California sea Hons Zalophus  californianus (Schusterman and Gentry 1971), Cape fur seals Arctocephalus pusillus (Rand 1955, 1959; Kastelein  et al. in press), Antarctic fur seals A. gazella (Bonner  1968; Payne 1979), and Steller sea lions  Eumetopias jubatus (Scheffer 1945; Fiscus  1961; Thorsteinson and Lensink 1962; Bryden 1972; Olesiuk and Bigg 1990). In all cases it appears that the blubber thickness of mature female otariids increases only slightly with age, unlike mature males where there is a marked change. Mature male otariids are larger on breeding sites than away from them (Scheffer and Wilke 1953; Thorsteinson and Lensink 1962; Lander 1979a; Calkins and Pitcher 1982). Studies of captive male otariids indicate that animals voluntarily regulate their food intake (Kastelein et al. 1990, in press) and increase body mass between 40 and 80% within a short 2 month period coinciding with wild populations approaching their breeding beaches (Schusterman and Gentry 1971; Spotte and Adams 1979; Ohata and Miller 1983). In addition to the fat which accumulates over all parts of the body (except theflippers),there also appear to be changes in core mass associated with increases in water content and/or proteinaceous tissue (Olesiuk and Bigg 1990). The animals presumably dehydrate and/or lose proteinaceous tissues during the 1-2 month fast and depend upon their accumulated body stores of water and energy to sustain them while defending breeding territories (Peterson and Bartholomew 1967; Peterson 1968; Gentry 1970). The breeding males return to their original weight at the end of the season (Spotte and Adams 1979). 10.4.2  N o r t h e r n F u r Seals  The growth curves estimated for male northern fur seals (Fig. 10.2) do not show the seasonal fluctuations in the size of mature males because of incomplete sampling  Chapter 10  Growth Fluctuations & Migration  162  in their migratory range and the reduced numbers of males available for sampling caused by the commercial male harvest. What I believe is shown is the core body size of mature males, prior to spring fattening. The change in the growth rate of male body mass between the ages of 4 and 5 y corresponds with production of sperm and onset of puberty (Scheffer and Wilke 1953). While male body mass increases linearly with age beyond puberty, there is a decrease in the annual length increment (see Fig. 10.2). The growth spurt reported by Scheffer and Wilke (1953) and Lander (1981) between the ages of 7 and 8 y is presumably related to the fattening phenomenon and not growth. Johnson (1968) found the age composition of 249 territorial males, and 156 adult males, that died of natural causes, ranged between 7-17 y (70% were aged 10-13 y; the modal age was 10 y). Based on the weight of testes, males reach their prime condition in about their 10th year (Scheffer 1950). Interestingly, adult females also appear to be in their reproductive prime at age 10 y, based upon the size of their fetuses (Chapter 2). The mean weight of males 10-13 y is between 110-150 kg (from Fig. 10.2). Territorial males killed and weighed within one week of arrival were 198 kg (n = 180).  1  Thus it appears that male body mass increases between 26 and 72% prior  to the harem bull arriving on land to defend breeding territories. This is in keeping with studies of captive male northern fur seals that increased in body mass by 40 to 80% and subsequently returned to their original mass the following fall (Spotte and Adams 1979). Captive male and female northern fur seals appear to voluntarily reduce their food intake from December to May (Spotte and Adams 1979, 1981; Ohata and Miller 1983; NPFSC 1984b). Weight gains occur from May onward with a marked synchrony between individuals. In the wild the weight gain appears to occur until 'calculated from Scheffer and Wilke (1953: £ = 188 kg, n = 12), Lander (1979a, weighed in 1965: 2 = 191 kg, n = 99), and Gentry and Holt (pers. comm., cited by Lander 1979a and weighed from 1975-77: x = 209 kg, n = 69)  Growth Fluctuations & Migration  Chapter 10  163  the seals arrive on land to breed and moult (Figs. 10.11 and 10.10). Lactating females haul out about 20-30% of the days between July-October to nurse pups (Bartholomew and Hoel 1953; Peterson 1968; Gentry and Holt 1986). Immature males haul out about 19% of the days (Gentry 1981) and appear to fast for at least part of the time. Kenyon (1956) found that less than 0.06% (27 of 57 239) of the young males taken in the commercial harvest had eaten prior to roundup. The average non-pregnant female in her reproductive prime weighs approximately 35 kg (Table 10.1). A mature male is about 3.4 times heavier than the female prior to mating, and is 5.4 times heavier when defending a breeding territory. High sexual dimorphism such as this is apparently positively correlated with the degree of polygamy of a species (Ralls 1977; Alexander et al. 1979). 10.4.3  Indeterminate G r o w t h  Traditional growth curves such as the Richards, Gompertz, logistic, and von Bertalanffy models (Zach et al. 1984) were not fit to the fur seal data because of the realization that growth was not monotonic, but in fact had strong seasonal components that should not be ignored. Secondly, some standard growth models are based in part on the premise that final body size is asymptotic, which does not appear to be the case for northern fur seals. The large body of data I have shown here suggests, rather unexpectedly, that growth in northern fur seals, and possibly in all other pinnipeds, is indeterminate (i.e. growth continues throughout the life span of the individual ). As shown in Table 10.1 and Figs. 2.7, 10.7 and 10.8, the annual 2  growth increment, which decreases with age, never reaches zero before the animal dies. Many species of fish, amphibians and reptiles, which live in the supporting medium of water, continue to grow beyond the self retarding plateau phase (Batt definition of indeterminate growth as defined by Lincoln et al. 1985  Chapter 10  Growth Fluctuations & Migration  164  1980). According to Bryden (1972) indeterminate growth occurs in these animals whose body temperature can vary with the environment (i.e. poikiliotherms). As such, some have concluded that growth of marine mammals is determinate (i.e. growth is limited during life span of an individual so that organism reaches maximum size after which growth ceases ) (see Bryden 1972). 3  Energy allocations between growth and other activities may explain the indeterminate growth observed in northern fur seals. A related factor influencing final body size is body form and the effect of gravity on supporting parts of the body. Marine mammals are fusiform with small appendages, living for most, if not all of their lives, in an aquatic environment (Laws 1956). Thus theoretically, marine mammals, like fish, may also grow indeterminately. Further evidence of indeterminate growth in marine mammals might be found by re-examining some of the historical pinniped data bases (e.g. harbour seals: Naito and Nishiwaki 1972) or by considering data from other species such as porpoise and dolphins. 10.4.4  Fluctuations in B o d y Length  It has long been recognized that length is a more reliable indication of pinniped size than is weight because it varies less (Scheffer and Wilke 1953; Laws 1956) and does not presumably fluctuate. The discovery of seasonal decreases in body length of fur seals is therefore mystifying. The large samples of seals measured negate the possibility that the reported seasonal change in length is an artifact of sampling location or error in body measurement. Nor is it due to yearly changes in growth rates and sampling locations. Periodicity in length was evident for samples collected during 3 year groupings 1958-62, 1963-68, and 1969-74 (Chapter 11). A decrease in body length implies a decrease in the length of the vertebral column. Since the seasonal change is of a relatively large magnitude (7 to 8%), it definition of determinate growth as defined by Lincoln et al. 1985  Chapter 10  Growth Fluctuations & Migration  165  is unlikely to result from deposition and resorption of bone in the vertebrae and intervertebral discs. Instead, it is probably related to factors that influence the spacing between the vertebrae discs. Three explanations worth considering are 1) changes in body water composition, 2) gravity and the astronaut phenomenon, and 3) the displacement of body mass when out of water. Bigg (1979) suggested there is a seasonal variation in water content of the cartilage and connective tissues separating the bony components. The higher the water content, the more edematous the soft tissue, the longer the vertebral column might be. For example, the rapid increase in body length of females in late pregnancy followed by a rapid decline after parturition could be explained by the effects of pregnancy hormones, progesterone and estrogen, which increases extra cellular water content (Guyton 1961). Following parturition, the hormone levels would rapidly decline and cause a loss of body water. Support for this view is contained in Taylor et al. where it is reported that the average right upper canine teeth of pregnant females weighed more than teeth from non-pregnant females. The difference in tooth weights is presumably related to water content. Thus pregnancy hormones could explain why pregnant females are longer than non-pregnant females of the same age. The second hypothesis concerns the effect of gravity. Like astronauts that become physically taller while in space, the length of a seal's body might expand while supported in water, and contract during its residency on land. Thus seals collected near the Pribilof Islands might be longest after completing their annual migration (fur seals are at sea the entire time), and shortest when leaving the Pribilofs to begin their fall migration. Finally, changes in length could be an artifact of mass displacement, exerting stress upon the vertebrae column. Remember that all the seals were measured on their backs on a 'cradle' measuring board. This means that the seal's body mass probably pushes down with considerable force upon the vertebrae column. The  Chapter 10  Growth Fluctuations & Migration  166  heavier the seal, the greater the force and the longer it would be. Thus seasonal lengthfluctuationsmight be a reflection of the seasonal weightfluctuations.Support for this hypothesis comes from measurements of fur seals on their backs and bellies, made by Yoshida and Baba (1981). They found that 85% of the seals were longer when the animals were lying on their backs. The increase averaged about 2% of body size (based on 10224 female and 119 male measurements). I believe that all three factors contribute to the seasonal fluctuation in body length. Seals undoubtedly retain more body water as their weight increases on their return journey to the Pribilof beaches in preparation for fasting on land, or as a direct result of pregnancy and the nutritional requirements of the fetus. However, the effect of fluid retention and gravity on the spacing of the vertebrae discs is probably of minor consideration compared to how body mass could alter the length of a seals measured on their backs. The possibility that mass displacement alters body length, begs questioning the standard procedure established by the American Society of Mammalogists (1967) of measuring seals on their backs. Perhaps seals should be measured belly down instead. The phenomenon of seasonal decreases in body length could be present in other species of wild mammals, but has not been detected because of insufficient sample sizes. It could also be missed because researchers fail to consider length changes as a real possibility. Note for example when Olesiuk and Bigg (1990) considered seasonal changes in the length of male Steller sea lions, they found a tendency for the animals to be longer in the spring than in the fall (adults were measured belly up). Finally the observation that pregnant female fur seals are longer than non-pregnant females may also be true of other species and is deserving of further attention.  Chapter 10  10.4.5  Growth Fluctuations & Migration  167  G r o w t h and Feeding Location  The seasonal fluctuation in the size of immature northern fur seals is related to the timing of migration and location of feeding. Seasonal growth patterns of mature animals appear to be more related to the reproductive cycle than to migration. Growth spurts and weight gains in all age groups only occurred when the seals were north of Washington coastal waters. The fact that animals either lose or fail to gain body mass while in Washington and further south may mean that the quality or abundance of food here is poor.  On the other hand, changes in body mass  may be inherent. Remember that captive seals appear to voluntarily reduce their consumption which may also be true of their wild counterparts. Perhaps weight loss and voluntary reduced consumption are evolutionary adaptations to lower seasonal abundances of prey encountered during this portion of migration. They could also be a mechanism to pace physiology for a long migration. Based on ocean circulation and distributions of commercially caught marine fish, the migrational range of the fur seal in the northeast Pacific Ocean can be divided into three major production domains: 1) the Central Subarctic Domain (Alaska), 2) the Central Coast Downwelling Domain (northern British Columbia), and 3) the Coastal TJpwelling Domain (southern British Columbia to Baja California) (Ware and McFarlane 1989). Fur seals gain body mass while in the first two domains. They lose it while in the third. Fur seals prefer to feed upon small schooling fish (Kajimura 1985; Perez and Bigg 1986). In the TJpwelling Domain (where seals are unable to maintain body mass), fur seals feed primarily upon northern anchovy when off of California. As seasonal availability of anchovy declines, the seals move along the coasts of Oregon and Washington, feeding upon a large variety of species including pacific hake, jack mackerel and Pacific herring. The greatest diversity of prey are consumed in  Chapter 10  Growth Fluctuations & Migration  168  Washington waters, with no single species dominating the diet. As the seals continue to migrate north into the Downwelling and Subarctic Domains, they begin to grow and gain body mass by feeding primarily upon Pacific herring (British Columbia, Gulf of Alaska, Western Alaska), Pacific sandlance (Gulf of Alaska), and capelin (Western Alaska, Bering Sea). Their diet changes mainly to walleye pollock when the seals enter the Pribilof Islands region. Washington waters are clearly a major transitional zone in the migration of northern fur seals (see Fig. 10.10). It is surprising that immature animals should spend so long here given that their growth is impeded until they leave this region. Perhaps water temperatures restrict the distribution of seals and therefore retards the northward migration of young animals. Energetic costs of heat regulation in colder northern waters may more than offset any gains made from going north earlier versus staying put. Another possibility is that there are spatial and temporal differences in prey and in theforagingefficiency of seals of different ages and sizes. By spreading out the migration, the total biomass of fur seals in any one area is probably never very high which would reduce competition for food among fur seals. Upon arriving at the Pribilofs, many individuals appear to fast and reduce their food intake. Furthermore, a large biomass of pollock in the Bering Sea can presumably support a large fur seal population and may well be the reason why large rookeries were established on the Pribilof Islands (Bigg 1990). 10.4.6  Future Considerations  Analysis of regional differences and temporal changes in growth rates should focus primarily upon young animals because of their high rate of development and sensitivity to food abundance. Seasonal fluctuations in body size appear to be more pronounced in immature fur seals than in mature individuals. The same may be true among phocids. Thus analysis of growth rates can be confounded by seasonal  Chapter 10  Growth Fluctuations & Migration  fluctuations unless appropriate precautions are taken.  169  Seasonal fluctuations can  also confound assessments of physiological condition based on length and weight measures. Population biomass may be a more useful ecological parameter than the number of individuals (Payne 1979). It is useful for estimating food consumption and the amount of energy required for growth (e.g. Innes et al. 1981; Fedak and Anderson 1982). It can also be used to assess the role of pinnipeds as energy consumers in the ecosystem. However, analyses which use estimates of population biomass should consider the strong seasonal and regional components of growth when assessing the amount of food consumed. This is particularly true if biomass estimates are based upon samples taken at or near breeding sites when pinnipeds are at maximal seasonal body size. There is a tendency in pinniped studies to be concerned with feeding conditions immediately adjacent to breeding areas and to assume that conditions beyond are constant. The fur seal data show that rapid growth and weight gains occur during a brief 1-3 month period prior to arriving on land. Body mass is gradually lost during the rest of the year. Although the quality and quantity of prey is important in the Bering Sea, it is also critically important as fur seals grow and gain body mass in the Gulf of Alaska and off the coast of northern British Columbia. Food is also important for the maintenance of body mass off California. The importance of feeding areas beyond the Bering Sea has not been generally recognized and should be given further attention considering the current decline of the Pribilof population. The same is true for other pinniped species. There is a need to better understand the interactions between pinniped migration, food abundance, growth and energetics because of the continual conflict between seals and commercial fisheries.  Chapter 10  10.5  Growth Fluctuations & Migration  170  Summary  Growth curves are described for males, pregnant females and non-pregnant females using morphometric measurements from over 18 000 northern fur seals shot at sea between California and the Bering Sea. Seals of all ages experience seasonal increases and decreases in body mass and length. Rapid growth and weight gains occur during a brief 1-3 month period as the population migrates through the coastal waters of northern British Columbia and Alaska on their way to the Pribilof Islands. Body mass is gradually lost during the rest of the year while fasting on land and wintering along the coasts of Washington, Oregon and California.  Given the pronounced  seasonal change in the body size of mature and immature fur seals, the fitting of standard growth curves is an oversimplification that must be considered when deriving estimates of population biomass. The fur seal data further suggest that growth in this species, and perhaps for all marine mammals does not stop at an upper asymptote, but continues throughout their life spans.  Chapter 11  Interannual Variability in the Growth of Adults and Juveniles from 1958—74 11.1  Introduction  There was a significant decrease in body size of northern fur seals during the Pribilof population increase from 1912 to 1950 (Scheffer 1955). This decrease was believed to be a density dependent response to increased competition for food. Intraspecific competition probably remained high until the fur seal population declined through the 1950s and 1960s. During this time, growth of individual seals was likely enhanced. In this study, I use morphometric measures recorded from seals shot at sea over the period 1958 to 1974 to test the hypothesis that body size increased as the population declined. In addition to examining density effects, I also consider how body growth might have been influenced by large scale environmental factors. Northern fur seals of all ages experience seasonal increases and decreases in body mass and length related to the timing of migration and area of feeding (Chapter 10). In general, rapid gains in mass and length occur during a brief 1-3 month period prior to the seals arriving on the Pribilof Islands in July.  171  Interannual Variability in Growth  Chapter 11  172  Table 11.1: Number of males collected at sea by month from 1958-74.  Year 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1958-74  Jan 0 10 0 8 0 0 0 0 1 7 14 0 6 2 6 0 2 56  Feb 19 8 0 7 16 0 0 0 2 6 19 13 7 3 6 0 0 106  Mar 17 23 38 67 18 0 1 0 52 2 13 22 8 20 5 0 0 286  Apr 74 44 42 74 23 3 14 29 13 6 30 27 9 14 5 0 0 407  May 123 16 74 33 12 8 18 6 4 5 36 30 9 29 15 0 0 418  Jun 24 14 46 6 78 14 8 3 0 0 79 0 9 0 0 0 0 281  Jul 4 0 39 0 37 57 21 0 0 0 38 0 0 0 0 16 24 236  Aug 0 0 11 0 40 51 19 0 0 0 14 0 0 0 0 39 31 205  Sep 0 0 0 0 17 4 7 0 0 0 0 0 0 0 0 19 4 51  Oct 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  Nov 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0  Dec 0 0 0 0 0 0 0 0 0 1 0 0 0 2 0 1 0 4  Jan-Dec 261 115 250 195 241 137 88 38 72 27 243 92 48 70 37 75 61 2050  I try to control for seasonal effects by restricting growth comparisons to specific months and age classes. First, I construct growth curves for seals collected over three consecutive time periods. I then estimate the mean size of mature, non-pregnant females collected each year from 1958 to 1974. Finally I estimate the annual growth rate of immature females and determine the 'condition' of those sampled.  11.2  Methods and Results  The pelagic measurement data were collected by Canada and the United States as members of the North Pacific Fur Seal Commission from 1958-74. A description of the procedures used to collect the fur seals along the west coast of North America, and the biases inherent in the data set are discussed in Chapter 10. Numbers of males, pregnant females, and non-pregnant females captured each month during the pelagic research program are contained in Tables 11.1, 11.2 and 11.3. Exploratory data analysis was conducted with the S software package (Becker and Chambers 1984) and statistical analyses were completed using BMDP statistical  Chapter 11  Interannual Variability in Growth  173  Table 11.2: Number of pregnant and post-partum females collected at sea by month from 1958-74.  Year 1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1958-74  Jan 0 95 0 311 0 0 0 0 30 41 55 0 42 48 20 22 25 689  Feb  Mar  156 576 2 285 32 0 0 0 105 26 74 52 18 3 6 8 0 1343  338 207 172 104 3 0 0 0 127 0 6 88 67 88 38 0 0 1238  Apr 145 327 296 216 16 0 58 88 21 17 51 40 18 48 23 0 0 1364  May 397 35 368 26 47 12 111 62 3 5 75 33 54 28 47 0 0 1303  Jun 98 96 151 1 333 113 7 35 0 0 138 0 2 0 0 0 0 974  Jul 0 0 174 0 176 278 44 0 0 0 43 0 0 0 0 151 113 979  Aug 0 0 74 0 309 555 234 0 0 0 39 0 0 0 0 166 123 1500  Sep 0 0 0 0 109 19 24 0 0 0 0 0 0 0 0 139 16 307  Oct 0 0 0 0 21 0 0 0 0 0 0 0 0 0 0 0 0 21  Nov 0 0 0 0 0 0 0 0 1 2 0 0 0 0 0 0 0 3  Dec 0 0 2 0 0 0 0 0 0 45 0 0 0 11 2 3 0 63  Jan-Dec 1134 1336 1239 943 1046 977 478 185 287 136 481 213 201 226 136 489 277 9784  Table 11.3: Number of non-pregnant females collected at sea by month from 1958-74.  Year  Jan  Feb  Mar  1958 1959 1960 1961 1962 1963 1964 1965 1966 1967 1968 1969 1970 1971 1972 1973 1974 1958-74  0 33 0 135 0 0 0 0 19 32 46 0 20 40 34 11 9 379  58 184 0 108 14 0 0 0 89 19 81 57 16 14 4 1 0 645  139 134 90 147 10 0 0 0 171 8 33 102 75 74 33 0 0 1016  Apr 136 169 122 236 55 5 77 114 36 30 81 62 33 70 37 0 0 1263  May 201 46 203 76 15 18 124 66 18 27 62 98 109 78 54 0 0 1195  Jun 41 20 31 6 181 66 47 58 0 1 94 0 19 0 0 0 0 564  Jul 33 1 58 0 96 112 75 0 0 0 84 0 0 0 0 23 21 503  Aug 0 0 15 0 211 291 88 0 0 0 40 0 0 0 0 57 51 753  Sep 0 0 0 0 87 6 22 0 0 0 0 0 0 0 0 65 14 194  Oct 0 0 0 0 26 0 0 0 0 0 0 0 0 0 0 0 0 26  Nov 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 1  Dec 0 0 1 0 0 0 0 0 0 36 0 0 0 12 1 2 0 52  Jan-Dec 608 587 520 708 695 498 433 238 333 154 521 319 272 288 163 159 95 6591  Chapter 11  Interannual Variability in Growth  174  software (BMDP 1988). 11.2.1  growth curves  Growth curves were plotted for year groups 1958-62, 1963-68 and 1969-74, to compare changes that might have occurred over the period 1958-74. The data were grouped to reduce possible biases associated with differences that occurred from one year to the next in population age structure and incomplete monthly samples (see Tables 11.1, 11.2 and 11.3). The first set of growth curves were drawn for immature males and females (aged 1.5 - 4.5 y). The second set were for pregnant and non-pregnant females (aged 5.5 - 15.5 y). In both cases, the analysis considered morphometric measures from the combined months of January through April because little or no seasonal change in body size occurs during this period, which is prior to both the immature growth spurt and the rapid growth of fetuses carried by pregnant females (see Chapters 10 and 2). The time series of morphometric measures were smoothed by locally weighted curves (lowess, see Cleveland 1984) and compared visually. These curves indicate that the mass and length of immature males and females gradually increased with age (Fig. 11.1). The same appears to be true for pregnant and non-pregnant females (Fig. 11.2). Fur seals seemed to grow at a much slower rate during 1958-62 compared to subsequent year groupings. From 1963-68 to 1969-74, there was an increase in body mass, but virtually no difference in body length. Perhaps the similarity of the length curves in the later years reflects a physiological maximum growth response to abundant food. The difference in body mass between years is marked and presumably means that seals became fatter with time (see Fig. 11.2). The inferences drawn from the smoothed growth curves are weak. I therefore tried to linearize the curves to statistically compare growth rates. If no statistical difference was detected between the slopes, I compared the mean body size over the  Chapter 11  Interannual Variability in Growth  175  Figure 11.1: Growth curves for immature males and females (ages 1.5 - 4 . 5 y) collected from January through April over three time periods: 1958-62, 1963-68 and 1969-74. The length and mass data were smoothed by lowess (/ = 0.60).  Chapter 11  Interannual Variability in Growth  176  110 -  Figure 11.2: Growth curves for pregnant and non-pregnant females (ages 4.5 y) collected from January through April over three time periods: 1958-62, 1963-68 and 196974. The length and mass data were smoothed by lowess (/ = 0.60). +  Interannual Variability in Growth  Chapter 11  177  three year groupings by analysis of covariance. The relationship between length and age, and mass and age for immature females were linearized by a square root transformation of age (measured in days). Growth rates (see body mass, Fig. 11.1 bottom left panel) increased from one year grouping to the next (0.71 to 0.77 to 0.81 k g d - ) and were significantly different from -0  each other  (^2,1607  = 4.07,  5  P = 0.017). Body length growth rates did not differ  significantly between the three time periods  (^2,1607  = 0.39,  P = 0.677). The mean  body lengths (adjusted for differences in age) were 102.1, 104.0 and 105.0 cm for the grouped years  = 36.73,  (^2,1607  P < 0.001). Post-hoc comparisons indicate these  differences between years were highly significant. Between the ages of 1 and 5 y, male growth is nearly linear (Fig. 11.1, right panels). Growth rates did not differ significantly between the three time periods (length: -^2,372  = 0.21,  P = 0.809; mass:  means did (length:  F , 72 2  3  F ,372 2  - 18.55,  = 0.04,  P = 0.957), but the adjusted group  P < 0.001; mass:  F ,372 2  = 17.23,  P < 0.001).  Post-hoc comparisons reveal that immature males were significantly smaller in 195862 than in the later year groupings. The immature male seals became progressively longer and heavier with time (length: 100.5, 103.6 and 105.1 cm; mass: 19.1, 20.5 and 21.9 kg). Only the difference in length between 1963-68 and 1969-74 was not significant (1374 = 1.47, P = 0.14). The statistical tests support the conclusions drawn from the smoothed growth curves shown in Fig. 11.1 that immature fur seals grew faster and attained larger body sizes as the Pribilof population declined through the 1960s and early 1970s. The same conclusion is drawn from the growth curves for mature females in Fig. 11.2.  11.2.2  size of mature non-pregnant females  I compared the mean length and body mass of mature non-pregnant females by year of collection. I only considered seals aged 14.5 y because annual growth +  Chapter 11  Interannual Variability in Growth  178  increments and seasonal oscillations in body size are minor at this stage of their lives (Chapter 10). I further pooled the data from all months of a given year because the sample sizes in some years were so small (i.e. only six seals sampled in 1967 were older than 14 y). The mean length of seals captured from 1958-74 should not change much with time because changes in length are largely a deterministic growth process. Furthermore, the year classes sampled overlap from one year to the next, and are not completely independent of each other. Body mass, on the other hand, is probably a relatively plastic growth parameter that could vary from one year to the next depending upon the quality and quantity of food consumed. Fig. 11.3 supports the hypothesis that body length should not have changed significantly over the period 1958 to 1972. The coefficient of the linear regression fit to the raw data was not significant (.Fi 678 = 0.71, P — 0.40). But there was a t  positive change in body mass (F\ e78 = 5.17, P = 0.02) suggesting that the mature t  females became heavier through the 1960s and into the 1970s. 11.2.3  growth rates of immature females  The lengths and weights of females sampled before October were pooled for ages 1.5 to 4.5 y according to the year sampled. The morphometric measures were regressed against the age of the seals (measured in days) to estimate the annual growth rate. Confidence limits were determined for the slope (growth rate) of each regression according to Zar (1984). The analysis was done by year of collection, and not by year class because body size varies seasonally and the months sampled were not the same each year (see Table 11.3). For example, if 2 y olds were sampled in April before the growth spurt, and 4 y olds were sampled two years later in July after attaining their maximal seasonal size, the estimated growth rate would be positively biased. Growth rates  Chapter 11  Interannual Variability in Growth  179  Figure 11.3: Mean length and mass of mature non-pregnant females aged 14.5 y collected from 1958-74. The vertical bars are 95% confidence limits. The means were smoothed with locally weighted curves (lowess: / = 0.60). +  Chapter 11  Interannual Variability in Growth  180  by year class can only be calculated and compared if the months of sampling are standardized which was not possible given the pelagic sampling design used. Growth rates of seals sampled in any given month of the year should be the same despite the seasonalfluctuationsin body size. For example if body length measured in April is regressed against age, the slope (but not the intercept) should be similar to the regression coefficient for samples from another month, such as July. I tested this by analysis of covariance. Selecting any given year (e.g. 1962), I compared the mean size of the immature seals caught each month (i.e. June, July, August, and September) after adjusting for age and testing whether the monthly growth rates were significantly different from one another (BMDP 1988). I found the adjusted group means of the 1962 samples differed (length F^ze = 5.91, P = 0.001 and mass F 6 = 9.08, P < 0.001), but 3t23  could not detect a significant difference between the monthly growth rates (length -^3,236  = 0.95, P = 0.42 and mass F 3e = 1-03, P = 0.38). I repeated this for za  two additional years (1961:Feb-May; 1968:Apr-Jul) where I felt sample sizes were sufficiently large. Again the monthly growth rates did not differ, but the adjusted mean sizes did. Thus I felt it reasonable to compare the growth rates by year of collection because the slopes within a year did not appear to differ significantly. Fig. 11.4 contains estimates of the annual growth rates for immature females. There is some suggestion of a periodic change in growth rates for body length, which declined to their lowest value in 1962, peaked in 1968, then declined thereafter. Body mass growth rates also peaked in 1968 but show no signs of having changed through the late 1950s, early 1960s. I was concerned that the differences in growth rates between years might be an artifact of population age structure. Perhaps the estimated growth rates depend upon the relative numbers of 2, 3 and 4 year olds sampled. I checked this by taking Monte Carlo samples, consisting of equal numbers of each age group, and found that  Chapter 11  Interannual Variability in Growth  181  Figure 11.4: Growth rates of immature females. The lengths and weights of seals aged 1.5 to 4.5 y were pooled by year of collection, then plotted against their age (in days) and fit with linear regressions. The slope of each regression (the growth rate) is plotted with 95% confidence intervals. The time series of growth rates were smoothed with locally weighted curves (lowess: / = 0.40).  Chapter 11  Interannual Variability in Growth  182  the growth rates estimated by repeated sampling were virtually identical to those shown in Fig. 11.4. These results give a much finer resolution of changes in growth than do the three year-grouped growth curves drawn in Fig. 11.1. I believe the year-grouped growth curves failed to detect the slower growth rates in the early 60s and 70s because the number of seals sampled during these years from January to April were few or none at all (see Table 11.3). Note however that Fig. 11.4 deals only with growth rates and infers nothing about changes in mean body size. 11.2.4  condition indices  The physical condition of a single seal was quantified with a condition index (CI) defined as the ratio between recorded mass (M) and expected mass (M) where  M ci =  T  .  (11.1)  M The expected mass of mature non-pregnant females was estimated from the empirical relationship between body length (L) and mass of non-pregnant females: M = 6.08 x 1 0  -5  X 2  74  (from Chapter 10). The condition index of immature females was  calculated from the allometric relation M = 2.055 X 10 L -4  2 , 4 6 9  derived for seals  between the ages of 1.5 and 4.5 y (n = 2 753). The mean fur seal condition was calculated by year of collection and plotted in Figs. 11.5 and 11.6. In theory, mean condition over time should fall about a value of 1.0 depending upon whether conditions in any given year were 'good' (CI > 1.0) or 'bad' (CI < 1.0). However the figures suggest that age effects are far larger than year effects. The condition of females improves with age (Figs. 11.5 bottom panel, and 11.6), but does not change in a clearly discernible way with time (1958-74). A linear regression fit to the data over time for age 14.5 nonpregnant females in Fig. 11.6 +  suggests there was a general improvement in body condition  (F\fi78  = 4.29, P =  Chapter 11  Interannual Variability in Growth  183  1.15 -  Figure 11.5: Mean conditions of immature females by year of collection. Each data point indicates the age group (2: 1.5-2.5 y; 3: 2.5-3.5 y; 4: 3.5-4.5 y). In the top panel the alternating dashed and solid lines join the mean condition of each age group by year class. In the bottom panel, the mean condition is smoothed by age group (lowess / = 0.60). The top and bottom lines were fit to ages 4 and 2 y, respectively The center dashed line is for age 3 y.  Chapter 11  Interannual Variability in Growth  x  CD  1.3  -  1.2  -  184  c o  1.1  •D  c o  O  1.0  -  0.9  -  58  60  62  64  66  68  70  72  74  Year Figure 11.6: Mean condition of mature non-pregnant females aged 14.5 y collected from 1958-74. The vertical bars are 95% confidence limits. The positive relationship between condition and time is described with a simple linear regression +  (^1,678 = 4.29,  P = 0.04).  0.04). But the condition of immature females (Fig. 11.5) may have declined from 1963 to 1972. There is a tendency for mean condition indices of different age classes to be clustered within a year, but vary markedly between years (Fig. 11.5 top panel). Furthermore, condition appears to vary from year to year independent of the cohort.  11.3  Discussion  It was not possible to simply compare fur seal growth in the late 1950s (when the population was relatively high) to growth at low abundance during the 1970s. Straightforward analyses were prevented by inconsistencies in sampling from 1958 to 1974. For one thing, sampling occurred in different areas of the North Pacific in different years. In addition, the population age structure changed with time, and the months of sampling and the number of seals measured varied between years. This was further complicated by the large natural variability in body size and by  Chapter 11  Interannual Variability in Growth  185  the pronounced seasonal fluctuations that occur as a function of age. I cautiously avoided biases attributable to sampling design and have tried to make valid comparisons of annual differences in growth. Unfortunately, the results and conclusions are rather weak. The data suggest that fur seals were smaller during the late 1950s and early 60s than in subsequent years. I base this conclusion largely upon the growth curves and the annual estimates of condition index and body mass from mature females (Figs. 11.2, 11.3 and 11.6). Comparing the grouped growth curves from immature females with the annual estimates of growth rates (Figs. 11.1 and 11.4) further suggests that these animals also attained larger body sizes with time, despite fluctuating growth rates. Rapid gains in mass and length occur during a brief 1-3 month period as the fur seals migrate through coastal British Columbia, the Gulf of Alaska and the Bering Sea prior to arriving on the Pribilof Islands in July (Chapter 10). The exact path and timing of migration is a function of age and experience (Bigg 1990) and may mean that different components of the fur seal herd have differing degrees of success in obtaining food. Unlike the older experienced seals, the younger and rapidly growing animals probably have greater difficulty procuring food. As such, growth of young seals is likely to be particularly sensitive to changes in food abundance that might be a function of oceanographic conditions. Fur seals feed primarily on small schooling fish, concentrating upon Pacific herring, capelin, sandlance, pollock and squid in the Northeast Pacific (Kajimura 1985; Perez and Bigg 1986). But virtually nothing is known about the size of these fish stocks and how they might have varied with time. The largest commercial fisheries in the fur seal's migratory range have targeted herring and pollock. Since fishing for these species began in the early 1960s, the amount of fish caught has varied considerably, suggesting the stocks undergo major fluctuations over many years.  Chapter 11  Interannual Variability in Growth  186  Fluctuations in population size of fish may be caused by natural environmental conditions. For example, at the start of the Bering Sea pollock fishery in 1964, the size of the virgin population was only about 2 million t (Bakkala 1989). Based on cohort analysis, the abundance increased dramatically to 12.4 million t in 1971. The strong pollock year classes, which appear to occur every 11 y, are thought to be related to sea surface temperatures and solar activity (Fig. 11.7; Bulatov 1989). Water temperature is believed to directly affect the abundance of pollock, and influence herring primarily through changed predation pressure (Laevastu 1983). The strong periodicity of pollock abundance also appears in commercial catches of cod, salmon and Dungeness crabs (Figs. 11.7). The period of the cycles for all these species is similar to the periodicity in growth rate of immature female fur seals (Fig. 11.4) and the changes reported in the size of fur seal pups on the Pribilof Islands (Chapter 9). The combined data from cod, crabs, salmon, pollock and fur seals begs questioning whether ocean productivity in the North Pacific is a function of an underlying driving mechanism, such as water temperature (Fig. 11.8) and solar activity (Fig. 11.9). Since the 1700s, the number of spots seen on the sun has regularly cycled with a period of about 11 y (Eddy 1976). This measure of solar activity correlates with weather patterns on earth such as stratosphere air temperature, tropical sea surface temperature, sea level pressure, and winter storms in the North Atlantic (Brown and John 1979; Labitzke 1987; Labitzke and van Loon 1988; van Loon and Labitzke 1988; Tinsley 1988). Chemical reactions caused by ultraviolet light may affect cloud particles which affects the amount of atmospheric heating (Tinsley 1988). This in turn could change ocean temperatures and currents, and affect the concentration, dispersal, and transportation of eggs and larvae, as well as their predators and prey. Changes in fur seal growth rates and body size can be attributed to a number  Chapter 11  Interannual Variability in Growth  187  Figure 11.7: Fluctuations in pollock recruitment and commercial catches of salmon, Pacific cod and Dungeness crabs in the North Pacific Ocean. Annual landings of Dungeness crabs for Washington coast ports (Grays Harbor, Willapa Bay and Columbia River) are from Armstrong (1983). Total landings of Dungeness crabs and salmon in northern California fisheries are from Botsford et al. (1982). The Pacific cod landings along the British Columbia coast are from Gunderson (1983). The estimated number of 3-y old walleye pollock in the Eastern Bering sea and Gulf of Alaska are taken from Wespestad and Traynor (1990).  Interannual Variability in Growth  Chapter 11  188  1950  1955  1960  1965  1970  1975  1980  1985  1950  1955  1960  1965  1970  1975  1680  1985  1950  1955  1960  1965  1970  1975  1980  1985  MAXIMUM MOXTHLY AVERAGE nONTBEIIMC SCA SURFACE TCflPEKATUItE AWOHAL ICS BT OTHUNES AREAS MINIMUM MONIMLT  Figure 11.8: Maximum, average, and minimum monthly bottom temperature anomalies in three areas in the Bering Sea from 1950 to 1982 (Area 1, eastern Bristol Bay; Area 2, south-central Bering Sea; Area 3, continental slope region), (from Laevastu, 1983).  Interannual Variability in Growth  Chapter 11  189  250-  ac 200 3  8  150-  E = 100  /  / / /  0  1960  1970  1980  1990  Figure 11.9: Solar activity from 1955-85. (from Khen, 1989).  of factors such as changes in food availability related to population density through intraspecific competition. But changes in food availability can also be related to interspecific competition with commercial fisheries or to natural fluctuations of the ecosystem. In the case of the Pribilof fur seal, about 75% of the population decline from the mid 1950s to early 1970s can be attributed to the harvesting and sampling of females (York and Hartley 1981; Trites and Larkin 1989). The remainder of the decline was caused by a high mortality of young (Trites and Larkin 1989). Thus I think it can be safely assumed that food abundance should have been in excess of the needs of the seals during the years of pelagic sampling (1958-74). Commercial fishing in the North Pacific and Bering Sea was in its infancy during the 1960s and probably did not have a major impact upon the fur seals' prey at that time.  Instead it appears likely that natural fluctuations in fish abundance  may have had a much larger impact than many might realize. Thus, along with commercial fishing, the possibility of natural fluctuations in the ecosystem need also to be taken into consideration when assessing the amount of fish available to northern fur seals, and indeed other apex predators. This is particularly relevant  Chapter 11  Interannual Variability in Growth  190  when assessing what role the large commercial fisheries of the 1970s and 80s may be playing in the continued decline of the Pribilof fur seal.  11.4  Summary  Analysis of morphometric measurements of northern fur seals made during the pelagic research program from 1958-74 suggest a periodicity in fur seal growth rates that might be indicative of underlying, large scale environmental factors. It further appears from the data that fur seals attained larger body sizes as the population declined through the 1960s. The conclusions, albeit weak, are based upon the mean size of mature females, annual growth rates of immature females, and growth curves for males and females. Definitive conclusions about changes in growth are complicated by inconsistencies in sampling and by large natural variations in body mass and length. Nevertheless, the data suggest the influence of environmental conditions on pinniped growth should be given greater attention when evaluating the factors, such as commercial fisheries, that affect the abundance of pinniped prey species.  Chapter 12  Haulout Composition, Capture Efficiency and Escapement During the Harvest 12.1  Introduction  Since 1835, the annual land harvest of Pribilof fur seals has consisted almost entirely of immature males captured at haulout sites adjacent to the breeding rookeries (Roppel 1984, Scheffer et al. 1984). These harvested animals provided biologists with large samples to estimate pup production and natural mortality rates and formed the basis for making many inferences about the general status of the population (Chapman 1961; Lander 1975, 1979, 1981; Scordino 1985; Fowler 1987; Trites 1989). Although commercial harvesting was stopped in 1972 on St. George Island and in 1985 on St. Paul, the historic catch data continue to provide insight into the population dynamics of northern fur seals (Eberhardt 1981, York and Hartley 1981, Smith and Polacheck 1984, Fowler 1990b, Trites and Larkin 1989). Further knowledge will likely be gained from the subsistence harvest now conducted on St. Paul Island by resident Aleuts. This study was concerned with the number of males using haulout sites during the period of harvesting, and was designed to estimate the rate of escapement from  191  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 192  the harvest and the efficiency of sealers at capturing the fur seals present. Capture efficiency is defined as the proportion of hauled animals forcefully moved to the killing fields. Escapement is the proportion of seals that use a given haulout and elude the sealers on the day of harvesting. Capture rate was determined at two haulout sites on St. Paul Island over a 4 year period from 1980-83. This was based on direct observations of the number of males captured at the haulouts and of the number that eluded the sealers. In addition to estimating escapement, these observations provided information about the abundance and relative ages of the animals using the haulouts. They indicated changes in fur seal abundances both within the harvest season and between years. The study also uncovered possible interactions between environmental factors, the behaviour of the seals, and the efficiency of the sealers.  12.2  Background to the Harvest  The commercial harvest on St. Paul Island generally began on July 1 and finished 5 weeks later (Lander 1980). The 25 day season usually consisted of 5 rounds of the Island in which every major haulout area was harvested once each week. There were 14 major areas driven on St. Paul Island (Fig. 12.1) consisting of approximately 32 haulout sites, of which 27 were accessible to the sealers. Nonbreeding animals tend to congregate adjacent to their rookery of birth on isolated hauling grounds referred to as haulouts (Nagasaki and Matsumoto 1957). The haulouts consist primarily of sub-adult males aged 2 to 5 and some bulls aged 6  +  that are unable to hold a territory. The seals are gregarious on land but appear  to segregate themselves by size. In general, bulls haul out a bit separately from other seals near the water edge. Progressively smaller animals occur further inland. The commercial and subsistence harvests of the haulouts has begun between 5 and 6 am and has been conducted by Aleut native people. Small groups of expe-  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 193  Figure 12.1: Rookery and hauling ground sites on St. Paul Island, Alaska. The bottom panels are the detailed inserts for the Little Zapadni and Lukanin sites (adapted from Lander 1980).  rienced sealers approach the young males from the downwind side of the haulout. They move quickly and quietly among the rocks and sand along the beach, until within 50 to 100 meters of the main group of seals. The pace breaks abruptly into a run between the seals and the sea, accompanied by the whistles and claps of the sealers. A few of the bulls charge towards the sea as soon as they first spot the sealers. The majority of fur seals bunch together and move inland away from the sealers. The seals are herded a short distance to the killing field and are allowed to rest frequently to prevent heat prostration. During the drive some of the larger seals  Chapter 12  Haulout Composition, Capture Efficiency & Escapement  that exceed the upper size limit imposed on the harvest are separated and allowed to return to the haulout. Occasionally harvestable seals may escape with an oversized animal or be too tired to be moved and thus released. The seals are herded together on the killing field. Small pods of 5 to 8 animals are separated from the large group and moved towards 6 experienced stunners. The stunners are in charge of killing only those animals below the legal size limits, while the oversized are allowed to escape back to the sea.  12.3  Study Area and Methods  Observations were made at Lukanin and Little Zapadni haulouts on St. Paul Island during the commercial harvest from 1980-83. These sites were selected because of the ease with which the initial round-up could be observed and because they were relatively small with short drives. This ensured that the observer could keep all sealers in view and be able to record all seals that escaped. The animals were classified as bachelors if they were shorter than the regulated 49 inch upper size limit of the harvest (124.5 cm, tip of nose to tip of tail, belly down); otherwise they were defined as bulls. The observer and sealers relied on their experience and judgement to determine the size of the seals. Bulls were unquestionably big with manes while obviously small seals and those of uncertain size were classified as bachelors. At no point were seals measured, either prior to or after the kill. Each haulout was harvested 3 to 5 times at approximately weekly intervals from June 30th to August 5th over 4 years. Lukanin and Little Zapadni haulouts are two of 32 recognized haulout sites on St. Paul Island (Fig. 12.1). Both are adjacent to a rookery and are the only major haulouts for each rookery. These two sites are typical of haulouts, although it must be remembered that all haulouts have their own peculiarities that may depend upon geography, the relationship of the haulout to the rookery or behavioural responses  194  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 195  of the different age groups using a particular haulout. The Lukanin haulout is a rocky reef at the water edge. It extends 10 meters inland and becomes a steep grassy hill strewn with boulders. The haulout is narrow at the water edge and does not contain a fortification of bulls defending territories. Most fur seals haul out on the grassy hill away from the water. The Little Zapadni haulout consists of boulders at the water edge. It rises 3 to 5 meters then slopes downward to a flat muddy area strewn with rocks. The haulout is TJ shaped. Large bulls are most prevalent near the water edge in the boulder area and appear equally spaced. Bachelors congregate further inland. Sealers tend to avoid the slippery boulders during the harvest and run along the edge of the flat muddy area. This allows the bulls and a few bachelors in the boulder area to escape. The observer moved with the sealers as they ran to cut the seals off from the water.  He was familiar with the harvest and recorded the number of bulls and  bachelors that escaped on counters held in each hand. At Lukanin haulout, the observer was able to see all the seals that escaped during the round-up; but he sometimes missed seals (generally bulls) in the boulder area near the water of Little Zapadni. After the round-up, the seals were herded to the killing field. The number of bulls and bachelors that were rejected or escaped along the drive and at the killing field was recorded, as well as the number of bachelors killed. The data were later tabulated to obtain the numbers of bulls and bachelors present on the haulout at the start of round-up and at the killing field at the end of the drive. Bachelors at the killing field were recorded as harvested or escaped. The number of bachelors that escaped at the killing field over 4 years were few (58 of 8,386 captured bachelors) and are not considered further in this report. Similarly, the number of females captured and released was also insignificant (7 females compared to 8,386 bachelors) and is not considered further.  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 196 Little Zapadni  1  30  J u l yBO  6 1  8 2  8 3  Lukanin  8  0  8 1  8  2  8 3  Figure 12.2: Number of bulls and bachelors present prior to each harvest at Little Zapadni and Lukanin during the month of July from 1980-83. The solid lines join 4 points, the mean number of males present during each harvest season.  12.4  Results and Discussion  12.4.1  C o m p o s i t i o n of the Haulouts  There was a 45% decline in the number of bulls using the haulouts at Little Zapadni and Lukanin from 1980-83 (Fig. 12.2: Spearman rank correlation coefficient r = s  —.447, P < .10 and r = —.620, P < .01, respectively). During this period there B  was an increase in the number of bachelors using Little Zapadni (r, = .329, P < .20), but a decline in the number of bachelors at Lukanin, except during the last year of study 1983. On average there were approximately 300 males per day at Little Zapadni and 350 at Lukanin. The daily ratio of bulls to bachelors averaged 1:1.67 at Little Zapadni and 1:3.88 at Lukanin.  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 197  Figure 12.3: Standardized number of bulls and bachelors present prior to each harvest at Little Zapadni and Lukanin from June 30 to August 4 (see text for details). The data points are indicated with symbols (0-1980, 1-1981, 2-1982, 3-1983) and were smoothed with a locally weighted curve (lowess using 2/3 of the data to smooth at each X point).  I standardized the yearly counts of seals to remove the annual trend and decipher the seasonal (July 1 - August 4) changes in abundance (Fig. 12.3).  I did this  by assuming the numbers of seals (iV»-,t) counted on day i in year t were from a Poisson distribution (i.e., the data were counts). The standardized score was thus N- = y/hj (iV, — Nt)l\JNf, where Nt was the average number of seals counted t  )t  over nt observations.  The trend was estimated by a locally weighted smoothing  routine (lowess, see Cleveland 1979). The data showed no change in the numbers of bulls using Little Zapadni during the month of July (Pearson r = .236, P = .347) and only a slight decline at  Chapter 12  Haulout Composition, Capture Efficiency & Escapement  Lukanin (r — —.400, P = .090). Trends in bachelor abundance differed markedly. At Lukanin the number of bachelors increased throughout the month (r = .519, P = .023), while at Little Zapadni they declined (r = -.370, P = .131). There did not appear to be any functional relationship between the numbers of bulls and bachelors present at the haulouts when the data were considered on a yearly basis (Fig. 12.4). Nor was there any apparent chronological relationship between bull and bachelor densities as the harvest season progressed (indicated by letters). This lack of relationship might be explained in part by the few observations and high variability inherent in the counts. When the observations of 4 years were overlaid (right most panel of Fig. 12.4) and linear regressions fitted for each set of data pooled by round, an apparent relationship emerges at Lukanin: the number of bachelors per bull increases with each successive round. At Little Zapadni the relationship is opposite. With the exception of round c, there are fewer bachelors per bull with each successive round. These conclusions are consistent with the seasonal trends (Fig. 12.3). Migrating males return to the Pribilof Islands in decreasing order of age and size from June until mid-August (Bigg 1986). I therefore expect counts of bulls to remain constant during the harvest period because they returned prior to the commencement of harvesting and are not being depleted. Bachelors on the other hand are being depleted by the harvest but are returning in increasing numbers. This accounts for the increase at Lukanin but not at Little Zapadni. I believe the decline at Little Zapadni is due to a harvest depletion of the older bachelors and a fortification of the water edge by the bulls that restricts access to the younger bachelors. Studies on tagged animals in the early 1950s showed that some haulouts were unavailable to bachelors (Kenyon et al. 1954). In particular, haulouts at Reef and Tolstoi (see Fig. 12.1) were accessible to only 35% of the males tagged at the home  198  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 199  80  81  82  83  80-83  300  200  i 2 100 CD  "o CD .Q  0 300  E 2  200  100  0 0  400  800  0  400  800  Number of Bachelors  0  400  800  Figure 12.4: Number of bulls versus bachelors using the haulouts on the day of harvest at Little Zapadni and Lukanin from 1980-83. The plot symbols indicate the chronological sequence (rounds) in which the observations and harvest were made. All of the observations were superimposed in the last panel and a linear regression fit to the data according to round (labelled with circled letter: i.e. one model was fit to all the as, a second to the b's, and so on).  rookery. This was in sharp contrast to the homing of 70 to 90% of the seals on haulouts at North East Point, Polovina and Zapadni (not the same as Little Zapadni). Nagasaki and Matsumoto (1957) suggested that the low return of young animals to Reef and Tolstoi was due to their inability to compete with bulls for a limited amount of space. I believe this was also occurring at Little Zapadni. It is difficult to explain the abrupt changes in the numbers of seals hauled out from one week to the next (Fig. 12.2). Knowledge of the particular haulouts and weather conditions leads me to believe that a combination of factors are responsible for variation in the number of bachelors and bulls using the haulouts; most notably the harvest, inclement weather, and the interaction between different age classes. Some diminished bachelor counts could be due to bulls preventing smaller seals  Chapter 12  Haulout Composition, Capture Efficiency & Escapement  from using a particular area. In other cases, hard rain and/or strong winds the night before or the morning of the harvest drives both bulls and bachelors to the water, leaving few on land. A few large haulouts of bachelors can also be explained by the presence of seals that were unavailable for harvest the previous week due to inclement weather or because they had not yet completed their migratory return to land. 12.4.2  C a p t u r e at the Haulouts  As with the numbers of seals on land, the ability of the sealers to round up bachelors and cut out bulls varied greatly from week to week (Fig. 12.5). Again some variability in the percentage captured can be attributed to haulout patterns and weather conditions. If the seals occupied the higher areas of the haulout, rounding up the entire haulout was possible, while dramatic decreases in percentage of the haulout captured occurred when there were unobtainable seals near the water (indicated by asterisks). On these occasions the haulout of bachelors was usually small (shown by vertical bars at the top and bottom of each panel). At other times, inclement weather resulted in haulouts that were essentially bare. The different composition of bulls and bachelors on the two haulouts does not affect the ability of the sealers to capture and roundup the seals (Fig. 12.5). Of all the bachelors counted from 1980 to 1983 on the two haulouts, the sealers captured 92.7% ± 3.1 (95% CI). More bachelors were captured on Little Zapadni (95.0% ± 2.9) than on Lukanin (90.3% ± 6.0), possibly because the physical layout of Little Zapadni enabled the escape route to be more effectively cut off. However the difference between the two haulouts was not significantly different [t = — 2.24, P = .07]. On average, only 41.5% ± 8.9 (95% CI) of the bulls were captured and driven to the killing fields (these bulls were later released). This is low compared to the capture of bachelors, confirming that sealers either avoid rounding up bulls or cut them out  200  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 201  Daily  Little Zapadni  Yearly  Lukanin  Figure 12.5: Percentage of hauled bulls (top panel) and bachelors (bottom panel) captured by the sealers at Lukanin (dashed lines) and Little Zapadni (solid lines) from 1980-83. The total numbers of seals present at each haulout prior to the harvest is indicated by vertical bars at the top and bottom of each panel. Both the daily rate of capture (left panels) and the weighted yearly value (right panels) are shown. The asterisks indicate the dates when many of the seals were near the water edge and unobtainable by the sealers.  Chapter 12  Haulout Composition, Capture Efficiency & Escapement  during the drive. Significantly fewer bulls were captured on Little Zapadni (39.6% ± 20.8) than at Lukanin (43.5% ± 14.3) [t = -2.783, P = .032]. This might again be explained by the physical layout of the haulouts. As previously noted, many bulls at Little Zapadni are in the boulder area near the water edge and are unobtainable or avoided by the sealers. It might also be easier to separate bulls at Little Zapadni because fewer animals are rounded up. There is no indication that the sealers become more efficient at capturing bachelors or avoiding bulls as the harvest season progresses. The slight decline in the number of bulls captured throughout the season and the increase in numbers of bachelors (Fig. 12.3) is a reflection of their changing abundances on the haulout site and not an indication of the efficiency of the sealers selecting the desired size class for harvesting. The ability of sealers to capture bachelors appears to depend on the number of bulls using the haulout sites (Fig. 12.6) and on the prevailing weather conditions (see following analysis). Significantly more of the hauled bachelors were captured by the sealers when large numbers of bulls were also present at both Lukanin [t = 4.11, P < .001] and Little Zapadni [t = 2.14, P < .05]. I suspect that bull densities determine how closely bachelors can approach the water edge. At high densities, bachelors are further inland and are easily cut off by the sealers; but at low bull densities, bachelors can occupy the water edge and therefore escape more readily. No one factor seems to determine the numbers of bulls that will be present at a haulout site. The presence of seals on the haulouts may induce others to join them, causing densities to 'snowball'. This could account for some of the large variability we observed in the numbers of seals using the haulouts from one round to the next. Some authors have noted that fewer animals use haulouts following hard rains and/or wind (Gentry 1981, Griben 1979). I examined hourly weather data collected on St. Paul Island by the National Climatic Center, NOAA, but  202  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 203  Little Zapadni  0  100  200  Lukanin  300 0  100  200  300  Bulls Present Figure 12.6: Number of hauled bulls present versus the rate of capture of bachelors at Little Zapadni and Lukanin from 1980-83. The data were fit with an unweighted linear regression.  could find no consistent pattern that could account for the variability in animal abundance. Perhaps the numbers present due to weather were confounded by the subsequent arrival of younger animals as well as the harvests of previous days. The success of the sealers at capturing bachelors may have been influenced by wind. On two days in particular (July 30/81 and August 3/82) there was a high escapement from Lukanin despite large numbers of bachelors ashore. The only thing unusual about these days was a lack of wind. The relationship between wind speed (recorded at 7 am) and the percent of bachelors captured at Lukanin was significant [t = 2.56, P < .01] suggesting that more seals escaped when the air was still because they were alerted much earlier to the sealers' presence. 12.4.3  Escapement  The two study sites indicate that the sealers captured an average of 92.7% of the bachelors present on the day of harvesting and 41.5% of the bulls.  Escapement  on both haulouts was similar despite the different make-up of the two sites. It is reasonable to expect similar estimates at other haulouts on St. Paul Island. There was little change in the efficiency of the sealers through the 5 week harvest season  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 204  or over the 4 years of study. This is not surprising since the sealers are experienced and the method of harvest is traditional and has been practiced in essentially the same way for the past two centuries. Indeed it is likely that the efficiency of the sealers has remained consistently high over the past century. Any drop in efficiency was explained by lack of wind and an absence of bulls that affected the behaviour of the bachelors. The size of the harvest depends upon the number of males at the haulout, the effort devoted to harvesting specific sites, the efficiency of the sealers trying to capture them, and the restriction imposed by the length limit.  Conversely, the  number of seals that escape the harvest and enter the breeding reserve depends upon the size of the animals, the length of time spent on land, and their ability to elude the sealers during harvest. Gentry (1981) estimates that only 19% of the young males that have returned to the Pribilof Islands are on land at any one time. The present study indicates that sealers capture 92.7% of the bachelors present. Bigg (1986) estimates that 71-100% of the captured 3 year old seals and 35-98.8% of the 4 year olds were harvested depending upon the length restrictions imposed from 1956 to 1982. The product of these 3 sets of estimates (% present X % captured x % killed) imply that the kill on any particular day removes 12.5-17.6% of the 3 year olds that use a given haulout and 6.2-17.4% of the 4 year olds. In other words, more than 82% of the 3 and 4 year old seals that have returned to St. Paul Island and use a haulout site are either absent or escape the harvest on any particular day. Any attempt to compare this estimate of escapement with those of others should be done with care.  There are at least 6 different sets of escapement estimates  contained in the northern fur seal literature (Kenyon et al. 1954, NPFSC 1962, Nagasaki 1961, Chapman 1964, Gentry 1981, Lander 1981). Unfortunately none of these estimates are comparable because each author used a different working definition of escapement.  Chapter 12  12 A A  Haulout Composition, Capture Efficiency & Escapement 205  Future Management Considerations  This study offers useful insights into the factors affecting the numbers of seals ashore and the ability of sealers to capture them. Future management efforts to assess the status of the herd may consider capturing (and releasing) haulout animals to record morphometric measures or to count and read tags.  This is certainly a viable option considering the results of this  study showing that sealers are able to capture 93% of the bachelors ashore. Assuming that the probabilities of hauling out on any particular day are independent of each other, over 50% of the population could be captured in 4 days (i.e. proportion of population captured = 1 - [1 - 0.19 x 0.927] y ). da  s  Knowing the capture rate is also useful for reconstructing and estimating the size of the male component of the fur seal herd. This can be done if the daily kill (either the commercial or subsistence harvest) is a proportion of the total population size. In this case the number of seals alive on the day of harvest equals the number killed, divided by the joint probabilities of being on shore, captured, and within legal size limits. Thus the male component of the Pribilof population can be constructed by age class and used to gain further insights into the dynamics of the fur seal herd. I feel the ability to easily capture immature males is a potentially useful way of obtaining additional information to assist in the management of the fur seal population and should be given further consideration.  12.5  Summary  The ability of sealers to capture northern fur seals was observed at two haulout sites on St. Paul Island (Little Zapadni and Lukanin) during annual harvests conducted from 1980-83. Males using these sites were classified as bachelors if within the size limit of the harvest (less than 49 inches in length) and as bulls if longer. Bulls were  Chapter 12  Haulout Composition, Capture Efficiency & Escapement 206  more numerous on Little Zapadni than Lukanin and appeared to restrict the access of bachelors to the haulout. Lukanin contained the greatest number of bachelors, as well as the largest total haulout. The abrupt changes in the number of seals using each haulout from one harvest round to the next were likely caused by a combination of factors, most notably previous harvest levels, inclement weather, and the interaction between different age classes. The ability of the sealers to capture the bachelors was dependent on the numbers of bulls present at each haulout; the more bulls on land, the greater the capture rate of bachelors. Sealer efficiency dropped on the few occasions when the low numbers of bulls enabled the bachelors to remain close to the water edge. A decline in sealer efficiency was also detected at low wind speeds, presumably because the bachelors were better able to detect the approaching sealers. On average, sealers captured 92.7% of the bachelors and 41.5% of the bulls. Escapement on both haulouts was essentially identical.  Chapter 13  Changes in Abundance and Body Size in Subadult Males 13.1  Introduction  In 1912, 5529 northern fur seal pups  (Callorhinus ursintis) were captured on the  Pribilof Islands, Alaska, and branded with the letter " T " on the top of the head (Lembkey 1913; Lander 1980a; Scheffer et al. 1984). For the next 10 years, males from this marked group were killed in the commercial harvest and measured to produce an age-length key. This study established that 3 y old males were between 104 and 116 cm long (measured tip of nose to base of tail, belly down, using a wooden caliper) and led to the regulated harvest of males between these lengths. Since 1918, the male-only harvest has been regulated by upper and lower size limits designed to capture 3 y olds. The original age length key was not re-examined until the early 1940s after a second group of branded pups began reappearing as subadults in the commercial harvests. The results of this experiment showed that a substantial number of 4 y olds were between 104 and 116 cm and indicated that the 1912 age-length standards no longer applied. This was confirmed by measuring branded and tagged animals from additional year classes, 1941 and 1947-49 (Scheffer 1955).  207  Chapter 13  Changes in Abundance & Body Size  208  Scheffer (1955) drew growth curves for two time periods, 1913-20 and 1941-52, and concluded that the fur seal decreased in size as the population rose from its all time low of 200 000 in 1910 to 1.8 million in the late 1940s and early 1950s. The decrease in body size was thought to be a primary result of increased competition among seals for food near the breeding grounds in summer. Subsequent work by Fowler (1990) confirmed a relationship between body size and population density. The goal of my study is to re-evaluate the response of body size to changes in population abundance as noted by Scheffer (1955) and Fowler (1990), using two sets of length measurements recorded from males harvested on the Pribilof Islands. The first data set contains the lengths of 1.6 million males of unknown age that were killed from 1933 to the late 1950s. The second data set contains the lengths of over 36 000 known-age males: those reported by Scheffer and Fowler, plus additional measurements spanning the period 1915-83. I begin with an overview of the commercial harvest and the methods used to collect and measure the animals. I then evaluate and correct for length biases associated with the duration of the harvest season, the enforcement of size restrictions, and the influence of tagging and marking on growth. Changes in male body size that occurred as the fur seal population increased and decreased are assessed, and implications regarding density dependent processes are discussed.  13.2  Background to the Harvest  The United States government hired Aleut natives to harvest northern fur seals from 1910-72 on St. George Island and from 1910-85 on St. Paul Island (Scheffer et al. 1984). Except for a brief period from 1956-68 when females were killed, the harvest has consisted uniquely of nonbreeding males that congregate adjacent to their rookery of birth on isolated hauling grounds referred to as haulouts (Nagasaki and Matsumoto 1957; Roppel 1984). The haulouts consist primarily of sub-adult  Chapter 13  Changes in Abundance & Body Size  males aged 2 to 5 and some bulls aged 6  +  209  that are unable to hold a territory.  Between 5 and 6 am, small groups of sealers ran along the waters edge of the haulout, cutting off the escape of the bachelors (Roppel 1984). The seals were herded a short distance inland to the killing fields. Once gathered into a large group, small pods of 5 to 8 animals were separated and moved towards 6 experienced stunners. The stunners were in charge of killing only those animals within the legal size limits, while the oversized or undersized were allowed to escape back to the sea. The duration of the harvest season has varied with time, beginning as early as June 1 and ending as late as mid August (Lander 1980a). Prior to 1930, there were sporadic hunts for food in October and November (Roppel 1984). Usually the harvest season consisted of 5 to 10 rounds of St. George and St. Paul Islands, in which every major haulout area was harvested once each round. In addition to the harvest season, upper and lower length limits were also varied as shown in Fig. 13.1 (Lander 1980a). From 1918-58, the harvest was restricted to seals between 104 and 116 cm (tip of nose to base of tail). The upper limit was gradually increased to 121 cm in 1961. In 1962, and all years hence, seals were measured from the tip of the nose to the tip of the tail, rather than to the base of the tail as was previously done. In all, the upper size limit was changed 12 times since 1959, ranging between 117 and 124 cm. In 1969, and in all years hence, the lower size limit was removed, while from 1963-71, the harvest included all oversized males without a mane (long silver colored guard hairs on the shoulders and on the back of the neck). Length was measured from 1918 to the late 1950s to provide an indication of the age of the animals killed, but more importantly, to impress visually and audibly on the clubbers the need to kill within the desired size limits (Scheffer and Kenyon 1948). Tally sheets of individual lengths indicate that every seal harvested on the Pribilof Islands was measured until 1956 on St. Paul and until 1958 on St. George.  Chapter 13  Changes in Abundance & Body Size  210  Figure 13.1: Body length limits used to control the harvest of male fur seals. Length is shown measured from the tip of the nose to the tip of the tail. The limits were adjusted to include the tail (Eq. 13.1) for the years prior to 1962. Note that the lower length limit was removed in 1969 and all oversized males without a mane were killed from 1963-71.  Thereafter, teeth were solely used to age the kill, and length was only recorded for known-age animals, generally those bearing tags.  13.3  Origins &: Implications of Sampling Biases  The lengths of seals collected in different years may not be comparable unless sampling conditions are understood and standardized. Since 1918, the size limits and methods of measuring have varied, and the duration of the harvest season has changed. These factors can influence the length estimates and may confound interannual comparisons. Similarly, the lengths of marked seals (tagged or branded) may not be comparable with the lengths of unmarked seals if marking has altered normal growth rates. Each of these factors is evaluated in turn. 13.3.1  M e a s u r i n g Techniques  From 1918-59, a wooden caliper (constructed by D.G. Hanna in 1915) was placed over the dead fur seal as it lay in the grass on its belly to measure the seal from  Chapter 13  Changes in Abundance & Body Size  211  the tip of the nose to the base of the tail, in inches (Scheffer et al. 1984). Standard length continued to be recorded in 1960 and 1961, but seals were now placed in a wooden 'cradle' and the length read from a tape fixed to the cradle, in both inches and centimeters (Abegglen et al. 1960). In 1962, the size limits of the harvest were modified to include the 2.54 cm tail (Lander 1980a). Written records contain no justification for the change in methods nor any indication that the estimate of tail length was based on field measurements. Harvested fur seals were now turned onto their backs in the 'cradle' measuring board. According to Scordino et al. (1984), the traditional technique involved pulling back the rear flipper until the tail was at the zero mark. Length was recorded in cm from the tip of the tail to the tip of the nose. The length of the tail must be added to the body lengths measured before 1962 to make the entire data set consistent. Data collected in 1983 (from the files of the National Marine Mammal Laboratory, Seattle) indicate that tail length varies with body size (measured in cm) according to t = -0.888 + 0.043 b  (13.1)  where t is the tail length and b is the body length measured from the tip of the nose to the base of tail (F = 22.74, n = 107, P < 0.001; Fig. 13.2). The data show that tails are 50% longer than previously assumed. Yoshida and Baba (1981) measured fur seals at sea by placing then in 'cradle' measuring boards with their bellies up and with their bellies down. They found that 85% of the seals were longer when the animals were lying on their backs. The increase averaged about 2% of body size and may be an artifact of the seal's body mass pushing down and extending the vertebra column (Chapter 10). It is not clear whether this bias extends to seals measured on the Pribilofs. But, it may be necessary to consider pre- and post-1962 sets of measurements separately.  Chapter 13  Changes in Abundance & Body Size  212  E x: © _ l  "co  90  100  110  120  130  Body Length (cm) Figure 13.2: Relationship between body length (tip of nose to base of tail) and tail length for 117 subadult males randomly sampled from the St. Paul commercial harvest in 1983 (data from the files of the National Marine Mammal Laboratory, Seattle, WA).  13.3.2  Size L i m i t Effects  Insight into the ability of sealers to kill seals within specified size limits are taken from two sets of observations made in the early 1950s and early 1980s. In 1950 and 1951, sealers were instructed to kill any seal with a 1947 tag, regardless of the 104-116 cm size limit (Kenyon et al. 1954). Following the harvest, all carcasses subject to the limits were carefully inspected for checkmarks indicating the seal had lost its tag. Both groups of animals, tag and tag lost, were measured. The distributions of lengths from tag and tag lost seals are compared graphically by box plots and empirical Q-Q plots in Figs. 13.3 and 13.4. They show the effectiveness of the size limits and indicate that 75% of the harvest was within the limits. The limits reduced the kill of undersized 3 y olds and oversized 4 y olds, but were not 100% effective. The data indicate that the harvest randomly sampled the 3 y olds in the population because few undersized animals were missed. However, the harvest was not representative of the 4 y old population because oversized individuals were more effectively excluded by the size limits.  Changes in Abundance & Body Size  Chapter 13  120  213  120 Age 3  RR §5 100  • 110 § (Q  3"  • 100 §  90 -  • 90  90  100  110  120  tag no tag  Length [no tag] Figure 13.3: Comparison of the distribution of measured lengths in 1950 from tagged 3 y olds with the lengths of individuals that lost their tags. The Q-Q plot in the left panel plots a scatter of the sorted values of the two data sets. The jittered data points should fall along the dashed line if the two distributions are the same. Boxplots in the right hand panel show the median and the four quantiles (each containing 25% of the data). Extreme data points are shown as outliers with dots. The notches on the medians mark 95% confidence limits and the horizontal dashed lines show the upper and lower size limits of the harvest (measured from tip of nose to the base of tail). All data are converted from inches to centimeters and were rounded at the time of measuring. The plots indicate that the lower limit prevented the killing of a few undersized 3 y olds, and that the 3 y old kill can generally be considered a random sample of the 3 y olds present in the population.  In 1980 and 1983, random samples of harvested seals were measured and aged by counting tooth annulli. The distributions of body lengths in Fig. 13.5 show that the animals harvested in 1983 were about 2 cm shorter than those killed in 1980. The lengths of all age groups are essentially normally distributed, except for a slight positive skewing in the lengths of 2 y olds in 1980. There is no indication of skewness in the length distributions of 4 and 5 y olds that should have occurred if an upper size Umit had been enforced. Indeed, over 50% of the 4 y olds should not have been killed in 1980 because they were oversized and exceeded the upper limits. In summary, the size limits imposed upon the harvest were better respected in the 1950s than in the 1980s. Undoubtedly the audible measuring of every seal killed  Changes in Abundance & Body Size  Chapter 13  130 £ g  130  Age 4  120 1  1  214  B  0  100  120 § CQ 110 | 100  100  110 120 130 Length [no tag]  tag  no tag  Figure 13.4: Comparison of the distribution of measured lengths in 1951 from tagged 4 y olds with the lengths of individuals that lost their tags. The data show that the upper length limit restricted the size of 4 y olds killed, and that the 4 y old harvest was not a random sample of the 4 y old population. Rest of legend as in Fig. 13.3.  until the late 1950s, impressed upon the clubbers the need to kill within the limits and improved their ability to size the moving seals. Once compulsory measuring stopped in the late 1950s, sealers appear to have selected seals according to age class rather than length, perhaps relying on visual cues such as whisker color and presence or absence of a mane. It seems unlikely that the fluctuating upper size limit (see Fig. 13.1) was reflected in the size of the animals harvested after 1960. This will not affect inter-annual comparisons of 3 y old lengths, since animals from this age group were too small to ever be affected by an upper limit. However the lax enforcement of the upper size limit will confound inter-annual comparisons of 4 y old lengths because of the progressive selection of larger 4 y olds with time. 13.3.3  Season Length Effects  Northern fur seals return to the Pribilof Islands each summer in order of decreasing age (Bigg 1986, 1990). The 4 and 5 y olds are among the first to arrive and account for a large proportion of the commercial harvest at the start of the season (Lander 1980a). As the season progresses, the age composition of the harvest changes. Har-  Chapter 13  Changes in Abundance & Body Size  • 1980  215  S 1983  150  90 2  3  4  5  2-5  Age (y)  Figure 13.5: Size distribution of subadult males harvested on St. Paul Island in 1980 (n = 682) and 1983 (n = 1824). The notches on the boxplots provide an approximate 95% test of the null hypothesis that the true medians are equal. The horizontal dashed line marks the upper size limit beyond which no seals should have been harvested. There was no lower size limit in either year.  vest statistics show the proportion of 3 y olds increases with time as the proportion of 4 y olds declines (Lander 1980a). The changing age composition of the harvest implies that the mean length of measured animals should decline throughout the harvest season. This was verified with the lengths of males harvested from 1933-58 and is illustrated with data from 80 000 seals killed in 1956 on St. Paul Island (Fig. 13.6). This figure shows large variability in the mean length of seals killed during the first two weeks of harvesting as the accumulated numbers of 4 and 5 y olds are depleted. As the season progresses, the harvest is increasingly made up of younger, recently arrived subadults. Hence the mean length of the total harvest (all days combined) declines with the length of the harvest season. Another concern is whether larger individuals of a given age class return to the Pribilofs before smaller ones. Insight into this was gained by comparing the  Changes in Abundance & Body Size  Chapter 13  £  O)  216  111.0  c  0) c  110.5  CD  110.0 80000 TO jg 60000 CO X  40000  o3 E  20000 -  3  0 -  T 20  1  1  30  10  June  1— 20  July  -1 30  1 10  r-  20  August  Figure 13.6: The influence of season length on the estimated mean size of males harvested on St. Paul Island in 1956. The bottom panel shows the cumulative number of males of unknown age that were harvested throughout the summer. The mean length of all seals sampled from the start of the harvest season to the given date are plotted in the top panel (measured tip of nose to base of tail). The dashed vertical line marks July 27.  lengths of 3 y olds by round in different years. Fig. 13.7 is representative of the 1  general pattern observed and shows the mean length of tagged and untagged 3 y olds collected over 7 rounds in 1963. There is no indication in this data set, or any of the others examined, of a trend in length with time. A similar conclusion was reached by Baker and Fowler (1990) after comparing tooth weights of harvested 3 y olds collected from 1948-84, over five 10-day periods. Thus it appears that the timing of migration is a function of age and experience, not of body size. In summary, the length of the harvest season will confound attempts to compare 'The sequence in which hauling grounds are visited to harvest seals. A circuit or round of the hauling grounds is usually completed in 5 days, and the procedure is repeated throughout the harvest of males.  Chapter 13  Changes in Abundance & Body Size  •  Untagged  •  217  Tagged  119 I)  111 2  3  4  5  6  7  Round  Figure 13.7: Mean length of tagged and untagged 3 y olds killed and measured over 7 rounds in 1963 on St. Paul Island (tip of nose to tip of tail). The bars are 95% confidence limits.  the size of harvested animals if age is not known. But variation in harvest season will not bias the lengths of known aged individuals. Thus the harvest seasons shown in Fig. 13.8 must be standardized to compare the mean lengths of the unknown aged seals harvested from 1933-58. Variations in the starting date of the harvest season are not important because of the depletion effect and the fact that all seals were measured each day of kill. This assumes that the seasonal return pattern is the same each year relative to calendar date which is supported by analysis of the daily age composition of the commercial harvest (Trites unpubdata). However, termination of the hunt must be standardized to the earliest completion date (July 27, Fig. 13.8) to prevent the inclusion of measurements from recently arrived, younger animals. 13.3.4  Tagging Effects  From 1957 to 1966 samples of tagged and marked pups consistently weighed less than untagged and unmarked pups (Roppel 1984). At the time, it was concluded that tagging and handling had caused a loss of weight and had slowed the normal rate of pup growth. This led to speculations that tagging might reduce future survival  Changes in Abundance & Body Size  Chapter 13  218  1960 -  1950 -  1940 -  W £  1930 1960 -  1950 -  1940 -  1930 30  15  30  June  15  July  30  15  30  August  Figure 13.8: Harvest days on St. Paul and St. George Islands from 1933-58. Note that no harvest was conducted in 1942. The dashed vertical line marks July 27 and is used in subsequent analysis to standardize the termination date of the harvest.  and growth. Until 1961, all known-aged animals measured were either branded or tagged. This was followed by a 3 year transition period, 1962-64, when both tagged and untagged animals were measured, after which all measured seals were unmarked and were aged from tooth annulli. Comparing the lengths of tagged and untagged animals may be invalid if tagging is indeed detrimental to growth. Fig. 13.9 compares the mean length of tagged and untagged 3 y olds captured from 1962-64. A two-way ANOVA confirms the highly significant difference in mean lengths between years (^2,7125 = 37.38,  P <  0.001).  However the difference between tagged and untagged individuals is not as significant as might be expected (JPI,7125 = 5.08, (interaction: ^2,7125 = 70.45,  P  = 0.026) and differs between years  P < 0.001). Post hoc pairwise comparisons, using the  Chapter 13  Changes in Abundance & Body Size  219  118 -  1962  1963  1964  Year Figure 13.9: Mean length of tagged and untagged 3 y olds killed and measured in 1962, 1963 and 1964 (tip of nose to tip of tail). Vertical bars are 95% confidence limits.  T U K E Y method (BMDP 1989), show tagged animals were significantly shorter than untagged animals in 1963, but were longer in 1964. There was no significant difference between the two groups in 1962. The difference between tagged and untagged 3 y olds in 1963 is broken down by rounds in Fig. 13.7 and shows the mean length of untagged animals was always larger (although not always significantly larger) than the mean length of tagged seals. A reassessment of the pup weight data from 1957-66 led to the conclusion that tagged pups grew at the same rate as untagged pups, but tended to be smaller at the time of tagging than average sized pups (Chapter 7). It appeared that tagged pups were born later in the breeding season and were more susceptible to being captured and tagged than older and heavier pups. The data shown in Figs. 13.7 and 13.9 further support the notion that seals were not randomly selected at the time of tagging. In summary, there appear to be real differences between the size of tagged and untagged animals. It is doubtful that the differences are caused by the presence or absence of a tag. Rather they are likely due to a biased selection of animals at the time of tagging. It seems that subadults bearing tags tend to be shorter  Chapter 13  Changes in Abundance & Body Size  —  St. Paul  220  -— St George  116  1930  1940  1950  1960  Year Figure 13.10: Mean length of males harvested until July 27 of each year from 1933-58 on St. Paul and St. George Islands (tip of nose to base of tail). The tiny vertical bars mark 95% confidence intervals. The upper size limit was removed in 1943 because no seals were killed in 1942. Lengths were converted from inches to centimeters.  than untagged animals, although this is not always the case (e.g. 1964). The difference between tagged and untagged seals is not clearly biased in one direction. Furthermore the size difference within a year may not be as important as the effects of inter-annual environmental conditions on growth. Thus it seems appropriate to compare tagged to untagged, although the potential for bias cannot be entirely ruled out.  13.4  Mean Length of Total Harvest, 1933-58  Measured lengths were pooled by year from the start of the harvest season to a standardized termination date of July 27. Table 13.1 and Fig. 13.10 compare the mean size of males harvested on St. Paul to those taken on St. George Island. Note that no seals were harvested in 1942 during wartime evacuation of the Pribilofs (see Fig. 13.8) and that the upper size limit was relaxed in 1943 to capitalize on the high 1942 escapement. In all other years, the harvest selected seals between 104 and 116 cm (tip of nose to base of tail).  Changes in Abundance & Body Size  Chapter 13  221  Table 13.1: Mean length, standard deviation (sd) and sample size of subadult males harvested until July 27 of each year from 1933-58 on the Pribilof Islands. Length was measured from the tip of the nose to the base of the tail and are summaries of daily tally sheets, converted from inches to centimeters.  Year "1933 1934 1935 1936 1937 1938 1939 1940 1941 1943 1944 1945 1946 1947 1948 1949 1950 1951 1952 1953 1954 1955 1956 1957 1958  St. George mean s3 n  lOlUl 109.44 109.07 109.12 109.77 109.90 109.08 109.44 110.76 112.97 109.43 109.13 109.86 110.40 109.43 110.07 110.12 110.30 109.82 109.93 109.41 110.26 110.56 109.90 109.37  3760 3.45 3.27 3.66 4.08 3.79 3.80 3.75 5.04 7.27 4.34 3.91 4.11 5.44 4.08 4.01 3.23 3.98 3.21 3.28 4.52 4.42 4.66 4.77 4.10  9365 9 821 10 337 7 571 9 251 10 545 12 765 13 359 10 876 18 538 6 580 14 632 7 977 9 270 11063 12 527 10 730 8 634 11 218 10 818 13 200 11872 13 309 6 243 8 548  mean  108771 108.21 108.84 108.72 108.08 108.30 109.66 108.93 109.43 115.40 109.68 110.10 110.32 111.42 108.61 109.51 108.58 109.36 109.86 109.31 110.35 111.18 110.31  St. Paul s3 n 3744 3.28 3.63 3.62 3.81 3.81 3.73 4.08 4.97 6.95 4.04 3.98 4.33 5.31 4.05 4.29 3.93 4.38 3.91 3.96 3.93 3.60 4.11  41711 40 488 40 782 38 072 40 639 40 814 47 547 49 263 58 469 72 969 31 162 44 661 35165 41942 52 195 53 646 46 015 44 674 48 340 46 089 47 343 37 555 48 658  Changes in Abundance & Body Size  Chapter 13  222  110 -  109  E u  108  -  O c  05  111  110  -  109  -  108  1930  1940  1950  Year  Figure 13.11: Inter-annual trend in mean length of males harvested until July 27 of each year from 1933-58 on St. Paul and St. George Islands (tip of nose to base of tail). Data from 1943 were removed before smoothing (lowess, / = 0.5). They show a general increase in the size of seals harvested since 1933. Locally weighted regressions (lowess) describe the overall pattern of the data in Fig. 13.11 (the 1943 harvest is excluded). On St. George, the mean size of the harvest increased from 1933 to 1958. There was also an increase in the size of animals harvested on St. Paul, but it occurred in two phases: 1933-47 and 1948-56. The sharp drop of 2.8 cm from 1947 to 1948 on St. Paul is puzzling. On St. George the drop from 1947-48 was only 1 cm and is within the range of natural variation. Certainly some of the annual variation in average size shown in Fig. 13.10 can be explained by shifts in the age composition of the population, which may even account for some of the differences observed between the two islands. However it is unlikely that age structure differed enough to account for the huge difference between the two islands from 1947-48. Instead much of the apparent drop on St. Paul  Chapter 13  Changes in Abundance & Body Size  223  was probably due to human error in measuring seals. Research records