INFORMATION TO USERS

This manuscript has been reproducedfrom the microfilm master. UMI films the text directlyfrom the original or copy submitted. Thus, some thesis and dissertation copies are in typewriter face, while others may be from any type of computer printer.

The quality of this reproduction is dependent upon the quality of the copy submitted. Broken or indistinct print, colored or poor quality illustrations and photographs, print bleedthrough, substandard margins, and improper alignmentcanadversely affect reproduction.

In the unlikely event that the author did not send UMI a complete manuscript and there are missing pages, these will be noted. Also, if unauthorized copyright materialhad to beremoved, a note will indicate the deletion.

Oversize materials (e.g., maps, drawings, charts) are reproduced by sectioning the original, beginning at the upper left-hand comer and continuing from left to rightin equal sections with small overlaps. Each original is also photographed in one exposure and is included in reduced form at the backof the book.

Photographs included in the original manuscript have been reproduced xerographically in this copy. Higher quality 6" x 9" black and white photographic prints are available for any photographs or illustrations appearing in this copy for an additional charge. Contact UMI directly to order.

U·M·I University Microfilms International A Bell & Howell Information Company 300 North Zeeb Road, Ann Arbor, M148106-1346 USA 313/761-4700 800/521-0600

Order Num.ber 9129695

The breeding ecology of the Laysan finch (Drepanidinae: Telespiza cantans) on Laysan Island

Morin, Marie Patricia, Ph.D.

University of Hawaii, 1991

V·M·I 300N.Zeeb Rd. Ann Arbor, MI48106

THE BREEDING ECOLOGY OF TRE LAYSAN FINCH

(DREPANIDINAE: TELESPIZA CANTANS) Ot~ LAYSAN ISLAND

A DISSERTATION SUBMITTED TO THE GRADUATE DIVISION OF THE UNIVERSITY OF HAWAII IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF

DOCTOR OF PHILOSOPHY

IN ZOOLOGY

MAY 1991

BY

Marie P. Morin

Dissertation Committee:

Sheila Conant, Chair Allen Allison Leonard Freed Robert Fleischer James Parrish G. Causey Whittow ACKNOWLEDGMENTS

My research on Laysan would not have been possible without the help of many different people and organizations. I am grateful to the Hawaii

Audubon Society and the Association of Field Ornithologists, whose grants funded part of this research. I am indebted to the U. S. Fish and Wildlife Service (USFWS) and National Marine Fisheries Service

(NMFS) for supplies, permits, transportation, and other support. I was privileged to work with many fine USFWS biologists and volunteers, especially Ed Bean, Stewart Fefer, Derral Herbst, Darcy Hu, Annie

Marshall, Larry Martin, Ken McDermond, Tim Ohashi, and Craig Rowland.

The monk seal people (NMFS) taught me how to "band" seal pups and also assisted me with finch censuses. My special thanks to Doris Alcorn,

Brenda Becker, Rusty Brainard, Mimi Brown, Pat Ching, Barry Choy, Tim

Gerrodette, Bill Gilmartin, Lisa Hiruki, and Robin Westlake. I spent many hours alone doing my research, but I thoroughly enjoyed spending my free time with these good people. I also thank Regie Kawamoto for her help in creating the maps for this dissertation.

I especially thank the friendly, professional officers and crew of the NOAA vessel, the Townsend Cromwell, which provided many trips to and from Laysan. Their excellent cuisine and hot, freshwater showers were much appreciated after several months in a field camp. lowe special thanks to crewman Dave, who fished me from the waters near Lisianski

Island after I fell overboard.

I am happy to thank the distinguished members of my doctoral committee: Professors Allen Allison, Lenny Freed, Rob Fleischer, Jim

iii

------Parrish, and G. Causey Whittow. Their many suggestions improved the different chapters in this dissertation, and their interest encouraged me.

I am most grateful to my committee chair, Sheila Conant, who facilitated this research in every possible way. She introduced me to those interesting "flying mice" on Laysan, and provided me with logistic, financial, and moral support. This research would have been impossible without her.

Last, but not least, I thank my husband, Joel Simasko. His easy­ going and helpful nature sustained me.

iv ABSTRACT

The breeding ecology of an endangered, endemic, non-migratory passerine was studied over three field seasons in the isolated, closed population that naturally occurs on Laysan Island in the Northwestern

Hawaiian Islands. Information on clutch size, egg size, chick growth, onset of breeding, nest placement and density, population size, and other variables was gathered on this little-known species, in order to understand the limiting factors currently operating on this population in its unpredictable environment.

The breeding data documented the high biotic potential of the

Laysan Finch. This, in conjunction with its omnivorous diet, helps to explain how the finch quickly rebounded from a severe population bottleneck in the early part of this century to achieve its current high density. Although these birds have delayed plumage maturation, and males generally do not breed until their third year, breeding birds can produce two and possibly more clutches per year. The typical clutch has three or more eggs, but a high proportion of these eggs disappear prior to hatching. A census during a year with abundant food resources showed that as much as a quarter of the post-breeding population may be recent

fledglings, but resightings of banded nestlings in later years suggested that very high post-fledging mortality may occur.

A small area immediately around the nest site (i.e. the nest

substrate) was defended, but a feeding territory was not defended and

foraging apparently occurred anywhere. This allowed active nests to be

close together, separated by as little as 2.2 m.

v Most of the vegetated area of the island had appropriate nesting

substrates; there appeared to be a highly specific preference for the

native bunchgrass, Eragrostis variabilis. Nesting occurred largely in

the two commonest vegetation associations: the Eragrostis association

and the Ipomoea-Boerhavia-Sicyos or viney association. Although the

viney association was less common, nest density was twice as dense in

that association. It was unclear whether nearby food resources or

microclimate considerations were more important in causing this

difference in density. Nest sites differed little between the two

associations.

In such a closed population, with no opportunity for immigration or

successful emigration, limiting factors must regulate the population within the carrying capacity of the island. There are no mammalian or

reptilian predators, but intra- and interspecific avian predation may be

a major limiting factor, mainly on eggs. Many of the eggs that disappeared probably succumbed to such predation. There were few nests

in which every egg hatched, primarily because such a high percentage disappeared, but eggs that hatched generally had a good chance of

producing a fledgling.

Females usually began incubation with their first egg, and eggs

hatched asynchronously in the order laid. Earlier-hatched chicks had more furcular fat at banding than did later-hatched chicks, and greater

furcular fat at banding was significantly associated with greater

fledging success. Except for nests where only two chicks hatched, earlier-hatched chicks had higher fledging success than did later­

hatched chicks. Several hypotheses for asynchronous hatching were

vi considered. No single hypothesis seemed to explain satisfactorily the observed data, although some predictions from the brood reduction, nest

failure, and insurance hypotheses may have been fulfilled. A more complex study might be helpful in narrowing the possibilities.

Overall, four-egg clutches fledged more chicks than two- or three­ egg clutches, but this varied among years, providing evidence for the hypothesis that the most productive clutch size in some years was not necessarily the optimal clutch size for all years.

In addition to intra- and interspecific avian predation, stochastic weather events emerged as a significant limiting factor, producing both direct and indirect effects. A severe storm during the wettest year directly affected the population by causing almost complete egg and chick mortality during an early breeding peak. Mortality was caused by nest flooding, chick starvation, and nest abandonment. Weather indirectly affected reproduction and recruitment through its effects on food resources. During the wettest year, food was abundant. During that year, larger clutches tended to have heavier eggs and had higher fledging success. During a drought year, the onset of the breeding season was delayed by more than a month, egg weights were significantly less than normal (the largest clutch size had the lightest eggs), and more malformed eggs were laid.

The limiting effects of weather (i.e. the drought) were verified in the following year by a decrease in population size. Three different methods suggested a real population decline: strip censuses, simultaneously-active nest densities, and nearest-neighbor nest distances.

vii Analyses of the visual strip censuses taken during pre-breeding, breeding, and post-breeding periods showed that individual counters were a significant source of count variability. The results also indicated the importance of knowing the stage of the breeding cycle at the time of the census; the population was apparently undercounted when most females were incubating and were therefore not seen.

Over 20 years of Laysan Finch population estimates suggest that rather large fluctuations are typical. The effective population size

(Ne) appears to be sufficient, at least for short-term maintenance, but whether the population is sufficient for long-term maintenance is uncertain.

In general, the breeding behavior of this bird had the major attributes typical of other cardueline finches, supporting its affinity with that taxonomic group.

viii TABLE OF CONTENTS

ACKNOWLEDGMENTS. ... iii

ABSTRACT •••• v

LIST OF TABLES • ... xii

LIST OF ILLUSTRATIONS. • xv

GENERAL INTRODUCTION •• 1 Literature Cited. • 8

CHAPTER I. THE BREEDING BIOLOGY OF AN ENDANGERED HAWAIIAN HONEYCREEPER, THE LAYSAN FINCH (TELESPIZA CANTANS) 10

Introduction. •• 10 Methods •••• 12 Study Site •••• 12 General Measurements • 14 Nest Visitation: Egg and Chick Measurements. • 16 Fates of Eggs and Chicks 18 Results •••••• 19 Weather. ••••• ••• 19 Pair Formation and Maintenance 19 Duration of the Breeding Season. • 22 Nest Situation and Description .• •••• 23 Within- and Between-Year Nest Site Fidelity. • 24 Nest Construction and Adult Behavior at the Nest •• 25 Egg Characteristics and Laying Interval. • 28 Clutch Size and Broods per Year. ••••••••••• 30 Duration of Incubation •••••••••••••••• 32 Hatchability, Hatching Success, and Hatching Order • 34 Nestlings, Fledglings, and Post-fledging Dependent Period. ••••• 35 Egg and Chick Mortality. • ••• 38 Discussion...... 41 Weather and Reproduction •• 41 Laysan Finch Breeding Strategy •••• 45 Mortality, Asynchronous Hatching, and Optimal Clutch Size •• 48 Comparison with Other Hawaiian Honeycreepers 54 Literature Cited. •••••••••.•••••• 56

CHAPTER II. LAYSAN FINCH NEST CHARACTERISTICS, NEST SPACING, AND REPRODUCTIVE SUCCESS IN TWO VEGETATION TYPES • 82

Introduction. ••••••• 82 Methods •••••• 83 Nest Characteristics 84

ix Nearest-neighbor Distances and Nest Density Estimates. •• 86 Clutch Size and Fledging Success 87 Results ...... 88 Nest Characteristics ••••••••••••• 88 Nearest-neighbor Distances and Nest Densities. 90 Fledging Success in Relation to Habitat Type • 92 Discussion. •••••••••••••••••• 93 Nest Characteristics ••••••••••••• 93 Nearest-neighbor Distances, Nest Density, and Fledging Success. • 97 Literature Cited. •• ••• 103

CHAPTER III. GROWTH AND MORTALITY OF LAYSAN FINCH (TELESPIZA CANTANS) NESTLINGS IN RELATION TO HATCHING ASYNCHRONY •• 121

Introduction. ••••••••• •••••• 121 Hatching Asynchrony Hypotheses • 122 Study Area and Methods. ••••••••• 126 Results •••••••• •• 128 Incubation Pattern •••••••••••• •• 128 Hatching Spread. • ••••• •••••• 129 Egg Weights in Relation to Laying Order. • 130 Overall Growth ••••••••••• ••• 131 Growth in Relation to Hatching Order ••• •• 132 Chick Fat in Relation to Egg Weight, Hatching Order, and Mortality •••••••• 135 Fledging Success Relative to Hatching order, Clutch Size, and Year. ••••••• 136 Discussion. ••••• •••• •••••• 138 Overall Growth • •• 139 Egg Provisioning • •••••• 140 Growth and Survival Relative to Fat Status, Clutch Size, and Hatching Order. ••• 141 Consideration of Hatching Asynchrony Hypotheses. •• 146 Summary ••••• ••••• 150 Literature Cited. •••• ••• 152 Appendix A. • 183 Appendix B. • • • • • ••••• • 185

CHAPTER IV. LAYSAN FINCH POPULATION ESTIMATES IN RELATION TO THE ANNUAL BREEDING CYCLE AND OTHER VARIABLES • •••••• 187

Introduction. • • 187 Study Area. ••••• •••••• 190 Methods •••• •• 192 Results •••••••• •••••• 196 Finch Distribution • •••• ••• 196 Comparisons Between and Within Years • 197 Estimation of Recruitment. •• • 198 Comparisons Among Individual Counters. •• 199 Comparisons Among Vegetation Types • •••••• 199 Evaluation of Transect Length. •••• ••• 200 x Survey Results from pre-1986 Counters. • 201 Discussion. ••••••••••••••••• ••••.• 201 Short- and Long-term Population Trends • ·• 201 Comparisons Among Individual Counters. • 205 Comparisons Among Vegetation Types ••••••• •• 207 Evaluation of Transect Length. ••••••••• •• 209 Assumptions and Evaluation of the Strip Transect Method. • 209 Management Implications and Recommendations • • 215 Literature Cited. ••••• • 219

GENERAL CONCLUSIONS AND COMMENTS •••• •• 233 Breeding Biology and Population Limiting Factors. •• 234 Population Estimates and Effective population Size. • 239 Phylogenetic Comments ••••••• 241 Conservation. •••• • 243 Future Research •••••••• • 246 Literature Cited. ••••••• 249

xi LIST OF TABLES

Table Page

1-1. Average Daily Maximum and Minimum Temperature and Average Daily Wind Speed •••..•.••••• •• •• • . 64

1-2. Abbreviations and Descriptions of Egg Fates. 65

1-3. Hatchability and Hatchling Survival for Laysan Finch Eggs from Known-size Clutches .••••••• 66

1-4. Development of Feather Tracts in Laysan Finch Nestlings. 67

1-5. Developmental Patterns in Young Laysan Finches • • • • • 68

1-6. Most Common Fates for Eggs from Clutches of Unknown Initial Size ...... 69

1-7. Fates for Eggs from Known-size Clutches in 1986 to 1988. 70

1-8. Most Common Fates for Eggs from Known-size Clutches for the Different Clutch Sizes •.••••••••••••••••.• 71

2-1. Primary and Secondary Substrate and Canopy Plants for Intensely Described Nests from 1987. •••••••••• 106

2-2. Number of Study Nests in Bunchgrass, Viney, and Mixed Vegetation from 1986 to 1988 • ••• 107

2-3. Frequency of Nest Orientation. • • 108

2-4. Nest Characteristic Variables and Definitions. • ••• 109

2-5. Summary of the 1987 Nest Characteristic Variables and Comparison Between Variables from Two Vegetation Associations •..••.• ••••••••••• • 110

2-6. Nearest-neighbor Distances for Simultaneously Active Nests •• 112

2-7. Density of Simultaneously Active Nests in the Study Area and Estimated Island-wide Active Nest Total at the Peak of Breeding •••••••..••••••••••••• 113

2-8. Comparisons of Mean Clutch Sizes between Two Vegetation Associations from 1986 to 1988 ••••• •••• • 114

2-9. Comparisons of Average Number of Fledglings per Nest between Two vegetation Associations from 1986 to 1988. •••••... 115

xii 3-1. Frequency of Clutches in Different Hatching Spread categories •••••• .••••••• • 157

3-2. Mean Weight Difference between the Last-hatched Nestling and the Largest Nestling for Various Clutch Sizes. •••••••• 158

3-3. comparisons of Measurements for Fledging and Non-fledging Chicks, Ages 9 and 10 Days Old Combined. •••• ••• 159

3-4. Richards' Growth Curve Parameters for Fledged Chicks of Different Hatching Order from Three-egg Clutches . . . . · 160 3-5. Richards' Growth Curve Parameters for Fledged Chicks of Different Hatching Order from Four-egg Clutches. .. · 161 3-6A. Comparisons of Measurements for Fledged Chicks of Different Hatching Order from Three-egg Clutches Where Two Eggs Hatched...... • 162

3-6B. Comparisons of Measurements for Fledged Chicks of Different Hatching Order from Three-egg Clutches Where Three Eggs Hatched. ••••••••••.•••••••••••••••• 163

3-7A. Comparisons of Measurements for Fledged Chicks of Different Hatching Order from Four-egg Clutches Where Two Eggs Hatched • 164

3-7B. comparisons of Measurements for Fledged Chicks of Different Hatching Order from Four-egg Clutches Where Three Eggs Hatched...... 165

3-7C. comparisons of Measurements for Fledged Chicks of Different Hatching Order from Four-egg Clutches Where Four Eggs Hatched...... 166

3-8. Number of Chicks at Banding Age with each Fat Status and Hatching Order •••••.••••• • 167

3-9. Hatching Success for Different Clutch Sizes. • 168

3-10. Percent of Hatchlings that Fledged Relative to their Hatching Order •• •• •••••••• ••• 169

3-11- Frequency of Number of Fledglings per Nest for Different Clutch Sizes, 1986 to 1988 Combined. · · · ···· · · ·· 170 3-12. Frequency of Number of Fledglings per Nest for Different Clutch Sizes, 1986 ...... · · · ·· · ·· · ·· 171 3-13. Frequency of Number of Fledglings per Nest for Different Clutch Sizes, 1987 ...... · · · · · ·· · ··· 172

xiii 3-14. Frequency of Number of Fledglings per Nest for Different Clutch Sizes, 1988 .••.•.•••• • 173

4-1. Laysan Finch Population Census Summary for 1986 to 1988. •• 224

4-2. Comparisons of Bird Numbers per Transect Among Three Vegetation Classifications during each Census and Coefficients of Dispersion. •••••• • •••• 225

4-3 Comparisons of Bird Numbers per Transect Among Individual Counters and Between Experienced and Inexperienced Counters • • • . • • • • • • • • • • • • • • • • 226

xiv LIST OF ILLUSTRATIONS

Figure Page

1-1. The Hawaiian Archipelago •• 72

1-2. Primary study Area on Laysan Island. • 73

1-3. Minimum Monthly Rainfall on Laysan •• 74

1-4. Temporal Distribution of Egg Laying Dates. 75

1-5. Distribution of Fresh Egg Weights during 1986-1988 76

1-6. Mean Fresh Egg Weights by Year in Relation to Clutch Size. 77

1-7. Distribution of Clutch Sizes by Year. 78

1-8. Incubation Lengths for 191 Eggs. ••• 79

1-9. Mean Number of Hatchlings and Fledglings per Nest for each Clutch Size in each Reproductive Period. ••••••••• 80

1-10. Fates of Eggs in Known-size Clutches by Year 81

2-1. Map of Laysan Island showing Primary Study Area. • • 116

2-2. Number of Simultaneously Active Nests in the Study Area during the 1987 Field Season •••••• ••• 117

2-3. Number of Simultaneously Active Nests in the Study Area during the 1988 Field Season ••••• • 118

2-4. 1987 Nest Map from Laysan Study Area • 119

2-5. 1988 Nest Map from Laysan Study Area • 120

3-1. Mean Fresh Egg Weights by Laying Order for all Clutch Sizes by Year. ••••••••• • 174

3-2. Mean Tarsus Lengths Plotted against Chick Age. . . • 175

3-3. Mean Beak Depths Plotted against Chick Age . . • 176

3-4. Mean Beak Widths Plotted against Chick Age . 177

3-5. Mean Beak Lengths Plotted against Chick Age. • 178

3-6. Mean Wing Lengths Plotted against Chick Age. • 179

3-7. Mean Sternum Lengths Plotted against Chick Age 180

xv 3-8. Mean Weights Plotted against Chick Age ••• .•• 181

3-9. Richards' Growth Curves for First-, Second-, and Third- hatched Chicks from Three-egg Clutches •• •••• 182

4-1. Vegetation Map of Laysan Island. •• • 227

4-2. Grid Map Used to Locate Transects on Laysan Island. • •• 228

4-3. Laysan Finch Population Estimates and 95% Confidence Intervals for 1986 to 1988 ••• •• •• •• • • • • • 229

4-4. Distribution of Finches per Transect for 1986 to 1988 Censuses ••.••••••••••• ••• 230

4-5. Tukey Groupings of Ranked Count Data • • 231

4-6. Laysan Finch Population Estimates and 95% Confidence Intervals on Laysan Island for 1968 to 1990. • • • • • • • • • 232

xvi GENERAL INTRODUCTION

Conservation biology is a "combination of art and science" (CPSU

1989) that draws from a number of traditional disciplines with the

common research goals of understanding dynamics of small populations,

identifying, studying, and preserving biodiversity, and restoring

damaged ecological systems. Indicator species in their native habitats

are often the focus of such research, even though the perpetuation of

the intact ecosystem itself may be of ultimate interest. The endangered

Laysan Finch (Telespiza cantans), which is the central focus of this

dissertation, is one such indicator species. It is one of the four

extant finch-billed Hawaiian honeycreepers (Family Fringillidae,

SUbfamily Drepanidinae; AOU 1983), an endemic Hawaiian subfamily of

birds. All the members of the tribe (Psittirostrini) are either extinct

(at least five fossil forms and three historically known forms; Olson

and James 1982a) or classified as endangered (USDI 1989). The Laysan

Finch itself has undergone a reduction in distribution. Paleontological

evidence indicates that, in addition to its current native distribution

on Laysan Island, it previously occurred naturally on at least two of the much larger Hawaiian islands (Oahu and Molokai), where it is now

extinct (Olson and James 1982b). The Laysan Finch has the largest

extant population of the Psittirostrini tribe, making this species a

good candidate for basic ecological research on the finch-billed

Hawaiian honeycreepers, despite legal restrictions on research due to

its endangered status.

1 The purpose of this dissertation research was to study the breeding ecology and population biology of this little-known species in relation to its environment. A better understanding of these complex ecological interrelationships will help resource managers to develop programs that will benefit the long-term survival of this species and the other finch­ billed Hawaiian honeycreepers in their natural habitats, as well as other insular species.

The Laysan Finch is a non-migratory species that has adapted to harsh environmental conditions on an uninhabited, desert-like island, and displays many attributes associated with such an environment.

Unlike some non-migratory insular birds, the finch is fully-flighted, although it commonly hops and runs short distances rather than flying

(Fisher 1903, pers. obs.). Like other species that have evolved in the absence of terrestrial mammalian and reptilian predators, the finch is curious and unafraid, approaching unfamiliar animals and objects with boldness (Munro 1960, Berger 1981). This fearlessness makes them easy to capture for research but also renders them vulnerable to predators and other causes of mortality.

There is no standing fresh water on Laysan (Ely and Clapp 1973,

Morin 1987), but a few shoreline springs feed directly into Laysan's large hypersaline lake and provide the finches with limited access to brackish water. The finches satisfy their water requirements during the driest months by eating plant parts (e.g. stems, leaves) with a high water content (pers. obs., S. Conant pers. comm.). During periods of extremely hot weather, they forage predominantly during early morning and late afternoon, avoiding the highest midday temperatures; when they

2 are not foraging they roost in the shade. Laysan Finches have heat tolerance similar to other passerines, but have reduced basal metabolic rate and evaporative water loss relative to other passerines of similar size (Weathers and van Riper 1982) - adaptations that are associated with dry, hot environments.

Prior to its designation as a national wildlife refuge, Laysan was privately leased for mining of the rich guano deposits left by its large seabird populations. The Laysan Finch had a brush with extinction in the early 1900's when rabbits, introduced by the guano miners, devegetated the island and caused the extinction of several endemic birds and plants (Ely and Clapp 1973). Before the rabbits were eliminated, the finch population entered a severe bottleneck that has left the population with very low heterozygosity (Fleischer et ale

1990). Unlike the other avian endemics, the omnivorous habits of the finch probably saved it from extinction. Without the plants, which are the mainstay of the finch's normal diet, a small population of finches survived, apparently by eating carrion, seabird eggs, roots, and old seeds in the sand. The population quickly rebounded after the vegetation recovered. However, since the size of the original natural finch population is unknown, it is not clear whether the current population size and the island's carrying capacity resemble the pristine condition.

In spite of its relatively large population of approximately 10,000 birds (USFWS 1984), the Laysan Finch is protected as an endangered species. Its highly localized natural distribution makes the population especially vulnerable to habitat destruction from hurricanes, tsunamis,

3 fires, and vegetation modification. Human activities, such as the accidental or deliberate introduction of non-native plants and animals, pose serious threats to the bird and its ecosystem. Because of their isolation, the finches are also extremely susceptible to diseases (Throp

1970, van Riper and van Riper 1985), some of which could potentially be transmitted to them by humans and equipment. Fortunately, many of these threats can be and are controlled with appropriate management.

Natural populations evolve in a complex milieu of inter- and intraspecific biological relationships and abiotic influences such as stochastic weather events and long-term climatic and geological changes.

On Laysan, the absence of terrestrial mammalian and reptilian (but not avian) predators, the present lack of passerine competitors, the lack of immigration and successful emigration, the simplicity of the ecosystem, and minimal disturbance by humans make it a uniquely suitable site for studying the basic ecological mechanisms that may influence and regulate a closed population of insular birds. An additional advantage is that relatively large numbers of individuals are repeatedly accessible.

This dissertation research focused on breeding biology and population dynamics, and examined some relationships between the two topics. In the course of research aimed at describing basic breeding biology, questions of optimal clutch size and hatching order were examined. The roles of weather and nest location were evaluated as possible population regulators by examining the relationship between these variables and reproductive success. Hypotheses, posed by others to explain asynchronous hatching, were examined relative to nestling growth and fledging success in nests where hatching order was known.

4 Finch populations were estimated directly by a standard (visual) line

transect method and indirectly by determining density and nearest­

neighbor distances among simultaneously active nests. Direct and

indirect estimation methods were compared and each was methodologically

evaluated. Finally, environmental data, finch reproductive success, and population estimates were considered together in order to draw

inferences about their ecological relationships.

Many variables in the breeding biology of a species determine its biotic potential: type of mating system, size of the nesting territory, number and size of clutches per year, hatching rates, fledging rates, chick growth, and ultimately recruitment. The dynamics of a population cannot be understood without some specific knowledge of these variables.

Chapters I and III explore the basic relationships among these life history parameters. Evidence for predation and stochastic weather events as possible population limiting factors are presented in Chapter

I.

Reproduction occurs in the context of the surrounding environment, which is not constant, but changes within and between years. Rainfall varied considerably among the three field seasons of this study; observations of finch food selection and plant phenology (Morin unpubl. data) indicated that changes in food resources corresponded to the variation in precipitation. Long-term studies on breeding biology are necessary if the range of environmental conditions and their sUbsequent effects on life history variables are to be observed and quantified.

For example, the puzzling fact that the seemingly most productive clutch size is not the most common size can be better understood when viewed

5 from a multiple year perspective. Chapter I examines the tradeoff between maximal production in a year of good food resources and consistent production in a year of average or poor food resources.

Like many passerines, Laysan Finches produce eggs that hatch asyn­ chronously. Females often begin incubation with the first egg, which causes eggs to hatch in the order laid over a relatively long time interval, rather than within a few hours. This produces clutches with chicks of varying sizes and ages. Over the past decade, many hypotheses have been formulated to explain the adaptive significance of this phenomenon. Chick growth and survival in general, as well as in relation to hatching asynchrony, were examined in Chapter III. In a harsh and unpredictable environment, such as that on Laysan, the brood reduction hypothesis seemed especially likely to explain such asynchrony, but closer examination of the data did not permit exclusion of other explanations. Several possible explanations for hatching asynchrony are discussed in Chapter III.

The lack of mammalian and reptilian predators, passerine competitors, and emigration opportunities mean that limiting factors other than these must be operational in this closed population. Changes

in population size and nest spacing over the three years of this study, monitored in concert with stage in the breeding cycle and fluctuations

in precipitation, provide evidence for the strong influence of

stochastic weather events on the breeding ecology and population dynamics of this species. These weather events may act directly on the population, or indirectly through their effects on available food

resources. For some other fringillid finches, food has evidently

6 assumed major proximate control over breeding (Newton 1973); this idea is not inconsistent with results of this research. Chapters II and IV examine nest spacing and population estimates as two different ways to evaluate year-to-year population changes, especially in relation to weather. These two methods of quantifying population size showed a population decline the year following the drought, which was consistent with some short-term reproductive characteristics (noted in Chapters I and III) observed during the drought year, such as changes in egg weights, date of onset of the breeding season, and shifts in the modal number of fledglings per clutch.

Variables that can confound passerine population estimates, such as stage of the breeding cycle when censuses occur, and effects due to different counters, are examined in Chapter IV. The feasibility and desirability of controlling some of these variables are considered, and recommendations for the timing and type of census are discussed in relation to its purpose.

7 LITERATURE CITED

AOU (American Ornithologists' Union). 1983. Check-list of North

American birds. 6th edition. Allen Press, Lawrence, Kansas.

877 pp.

Berger, A. J. 1981. Hawaiian birdlife. 2nd edition. The Univ. Press

of Hawaii, Honolulu, Hawaii. 260 pp.

CPSU (Cooperative National Park Resources Studies Unit). 1989.

Conservation biology in Hawaii. C. P. stone and D. B. Stone, (eds.).

Univ. of Hawaii Press, Honolulu, Hawaii. 252 pp.

Ely, C. A. and R. C. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll. Res. Bull. No. 171.

Smithsonian Institution, Washington, D. C. 361 pp.

Fisher, W. K. 1903. Birds of Laysan and the leeward islands, Hawaiian

Group. U. S. Fish. comm. Bull. 23 (pt. 3): 767-807.

Fleischer, R. C., S. Conant, and M. P. Morin. In Press. Genetic

variation in native and translocated populations of the Laysan finch

(Telespiza cantans). Heredity.

Morin, M. P. 1987. Laysan Finches drown as a result of marine debris.

'Elepaio 47: 107-108.

Munro, G. C. 1960. Birds of Hawaii. Revised edition. Charles E.

Tuttle Co., Inc., Rutland, Vermont. 192 pp.

Newton, I. 1973. Finches. Taplinger Publishing Co., Inc., New York,

New York. 288 pp.

8 Olson, S. L. and H. F. James. 1982a. Prodromus of the fossil avifauna

of the Hawaiian Islands. smithsonian Contrib. Zool. No. 365.

Washington, D.C. 59 pp.

Olson, S. L. and H. F. James. 1982b. Fossil birds from the Hawaiian

Islands: evidence for wholesale extinction by man before western

contact. Science 217: 633-635.

Throp, J. 1970. The Laysan finch bill in the Honolulu Zoo. 'E1epaio

31: 31-34.

USDI (U. S. Dept. of the Interior). U. S. Fish and Wildlife Service.

1989. Endangered and threatened wildlife and plants. 50 CFR 17.11

and 17.12.

USFWS (U. S. Fish and Wildlife Service). 1984. Recovery plan for the

Northwestern Hawaiian Islands passerines. U. S. Fish and Wildlife

Service, Portland, Oregon. 66 pp. van Riper, S. and C. van Riper III. 1985. A summary of known parasites

and diseases recorded from the avifauna of the Hawaiian Islands. Pp.

298-371. In Hawai'i's terrestrial ecosystems: preservation and

management. C. P. Stone and J. M. Scott (eds.). CPSU, Univ. of

Hawaii, Honolulu, Hawaii.

Weathers, W. w. and C. van Riper III. 1982. Temperature regulation in

two endangered Hawaiian honeycreepers: the Pali1a (Psittirostra

bai1leui) and the Laysan Finch (Psittirostra cantans). Auk 99: 667­

674.

9 CHAPTER I

THE BREEDING BIOLOGY OF AN ENDANGERED HAWAIIAN BONEYCREEPER,

THE LAYSAN FINCR (TELESPIZA CANTANS).

INTRODUCTION

The Laysan Finch, Telespiza cantans (Wilson 1890) is an endangered member of an endemic Hawaiian subfamily (Fringillidae: Drepanidinae) called the Hawaiian honeycreepers (A.O.U. 1983). This group of birds is known for its spectacular adaptive radiation, but unfortunately is noted also for its high proportion of endangered or extinct taxa (U.S. Fish and Wildlife Service 1984, Freed at al. 1987, Scott et ale 1988).

The breeding biology of Hawaiian honeycreepers in general is poorly known, although considerable progress has been made, especially in the past 30 years (Berger 1969, Berger et ale 1969, Berger 1970, Eddinger

1970, van Riper 1978, Scott et ale 1980, van Riper 1980, Sakai and

Johanos 1983, Collins 1984, Kern and van Riper 1984, Freed et ale 1987,

Freed 1988, Pletschet and Kelly 1990). Additional breeding studies are

in progress (S. Conant pers. corom., L. A. Freed pers. corom., J. Jacobi pers. corom.).

The breeding biology of the Laysan Finch has been little studied,

partly because of the remote location of its current distribution.

Laysan Finches occur naturally only on the remote, uninhabited Pacific

island of Laysan in the Hawaiian Archipelago; a small introduced

population also occurs at Pearl and Hermes Reef (Figure 1-1). Fossil

evidence confirms that Laysan Finches once occurred on Oahu and possibly

on Molokai (Olson and James 1982).

10 Generally, biologists have not had an opportunity to make prolonged studies during their short visits to Laysan Island. Ely and Clapp

(1973) summarized the literature on Laysan Finch biology; other information on breeding phenology and nest sites is scattered in older literature (Rothschild 1893-1900, Fisher 1903, Dill and Bryan 1912) and in unpublished trip reports (Kramer 1959, Woodside 1961, Conant 1985).

Unpublished government documents by Crossin (1966) and Sincock and

Kridler (1977) also contain information, primarily anecdotal, about

Laysan Finch breeding.

The intent of this research was to describe the basic breeding ecology of a poorly known endangered Hawaiian honeycreeper, with the special focus of assisting in the conservation of this species in its native habitat. Limiting factors such as stochastic weather events and predation (which in this case includes inter- and intraspecific avian predators, but no mammalian or reptilian predators) are of broad interest in relation to population regulation in any closed ecosystem.

Other closely related, endangered, finch-billed Hawaiian honeycreepers, such as the Palila (Loxioides bailleui) and the Nihoa Finch (Telespiza ultima) are much rarer than the Laysan Finch, and hence more difficult to study. It is hoped that information from this paper will also contribute to understanding the ecology of these rarer species, as well as insular passerines in general.

11 METHODS

Study Site

Laysan is located approximately 1,506 km northwest of Honolulu (25 0

46' N 171 0 45' W), and is part of the Northwestern Hawaiian Islands

National Wildlife Refuge (Figure 1-1). The island is about 2.9 km long and 1.7 km wide, with a maximum elevation of 10.7 m above sea level.

The central portion of Laysan is covered with a shallow, hypersaline lake (Figure 1-2). Although there is no standing fresh water, there are a few fresh water seeps along the lake shore (Ely and Clapp 1973, pers. obs.). Only about 187 (47%) of the island's 397 ha are covered with vegetation; the rest is open sand and lake. Because of its more northerly latitude, Laysan has a more temperate climate than the inhabited main Hawaiian islands. Summers are generally hot, and winter months usually much cooler, with more frequent storms. Although extreme weather conditions are probably more common during the winter months, heavy rain and high wind may occur at any time. Weather data for the

1986-1988 field seasons appear in Table 1-1 and Figure 1-3.

The original flora and fauna of this sand island were highly endemic (Lamoureux 1963, Newman 1988). Originally the fauna was composed of 17 species of seabirds, the green sea turtle, the Hawaiian monk seal, various migratory shorebirds, five species of endemic land birds, and numerous native terrestrial invertebrates (Ely and Clapp

1973).

There is no evidence that prehistoric Polynesians ever inhabited

Laysan. The island was mined extensively for guano starting in 1890, and although Laysan was named a bird refuge in 1909, guano was mined 12 there until 1910. Poaching for seabird feathers continued intermittently until at least 1915 (Ely and Clapp 1973). By 1903, rabbits had been introduced as a food source for the guano miners. The rabbit population increased quickly, causing almost total destruction of the vegetation, a catastrophe from which Laysan has never fully recovered. One of the three endemic plants, three of the five native land birds, and a number of terrestrial invertebrates became extinct

(Butler and Usinger 1963, Ely and Clapp 1973). The Laysan Finch was one of the two endemic bird species (the other being the Laysan Duck) that survived the ecological disaster, which ended in 1923 when the last rabbits were killed. The post-1923 Laysan ecosystem, although still mostly composed of native plant species, is different from the original ecosystem in several ways. Laysan Finches no longer co-exist with the three now-extinct land bird species, with whom they may have competed.

In addition, non-native plants and invertebrates have been introduced to the native flora and fauna over the past 100 years. These new species have undoubtedly affected the native plants and invertebrates (e.g. through competition and predation). Some introduced species may provide new food sources for the Laysan Finch, but weedy species (such as

Cenchrus echinatus) may threaten the regeneration of the primary substrate for finch nests, Eragrostis variabilis (Morin and Conant

1990).

Newman (1988) detailed eight vegetation associations on Laysan, which represent a more detailed breakdown of the five associations mapped by Lamoureux (1963). When mapping nests in my study area, the five associations I recognized were the same as Lamoureux's, except that

13 I omitted the ephemeral beach Nama association (where I seldom saw finches) and I added the Pluchea association, a non-native shrub association that apparently was not present on Laysan in 1961.

The five associations occur more or less concentrically around the lake, from outermost to innermost, as follows: the Scaevola shrub association; the Eragrostis bunchgrass association; the Ipomoea­

Boerhavia-Sicyos viney association; the Pluchea shrub association; and the sesuvium-Heliotropium-Cyperus wetland association.

From 1986 to 1988, representative areas from each of these vegetation types were searched for nests. In 1986 my nest studies were concentrated in the northern third of the island, whereas in 1987 and

1988 they were restricted to the northwestern portion of the island

(Figure 1-2). I concentrated my efforts in these limited areas not only to reduce time spent walking from one area to another, but also to reduce the likelihood of accidentally collapsing the burrows of nesting procellarid seabirds during extensive walking.

General Measurements

Three field seasons were spent on Laysan: 20 February to 3 August

1986, 7 April to 22 July 1987, and 14 May to 30 August 1988. For 5 or 6 months during each year, daily records of rainfall (rom), wind speed

(km/hr) and direction, and maximum/minimum temperature (degrees centigrade) were kept.

A total of 1,106 adult, subadult, and nestling finches were banded.

Most adult and subadult finches were captured either with an insect net or in baited wire Potter traps. In addition to morphological

14 measurements, finches were weighed to the nearest 0.5 g with a hand-held

50-gram Pesola spring scale, and age, sex, stage of molt, presence or absence of a brood patch, and status of the cloacal protuberance were recorded. Accumulation of fat in the furcular or interclavicular region

(the area between the attachments of the pectoralis muscles to the furculum and coracoids) was estimated on a scale of 0 to 4: 0 represented no fat present and the region was deeply concave; 1 represented traces of fat in the still deeply concave region; 2 represented some fat in the slightly concave region and on the clavicles; 3 represented fat filling the region nearly level with the pectoralis muscle; and 4 represented a convex pad of fat overflowing the length of the furculum.

Each bird was banded with a unique band sequence which included a numbered u.s. Fish and Wildlife Service (USFWS) aluminum band and plastic leg bands of different colors. The metal band was used as one of the "colors" in each four-band combination. Birds that were banded and measured were caught in the northern one-third of the island; most were from the immediate vicinity of the beach camp or from the primary study area along a trail between the west camp and the lake (Figure 1-

2) •

Observations of finch food selection and plant phenology were recorded in all three field seasons (Morin unpubl. data), but these data are not included here.

15 Nest Visitation; ~ and Chick Measurements

Nests were located by seeing one or more of the following clues: 1) a singing adult male, which indicated a nearby nest 2) an adult finch entering or leaving an ~. variabilis clump, 3) a female (often pursued by her mate) carrying nesting material, 4) a female begging for food from her mate, 5) one or both members of a pair staying abnormally close to me, which often meant I was very near a nest, and 6) a female late in the season who had not yet begun her post-breeding molt. Nests of unknown initial clutch size were sometimes discovered by hearing the chicks beg. When I was near a nest, finches would often stand tall on their toes and stretch out their necks while peering at me in a characteristic,. craning manner. Most nests were located fortuitously during random walks when feeding observations were being recorded. Some nests were found by methodically searching through Eragrostis clumps, without the benefit of any clues. Although this method was much less productive, I employed it especially during the 1988 field season, because I arrived on Laysan after the peak of nest building.

Each bunchgrass clump that contained a nest was marked with colored plastic ribbon. A numbered plastic flag attached to a metal wire was also placed near the nest, so that the flag was visible from the study area trail. In 1987 and 1988, maps of the nest locations were made to scale (see Chapter II). The plant substrates containing the nests were recorded in all years, as was the vegetation association where the nests were located. Nest substrates, locations, and characteristics are thoroughly detailed elsewhere and will not be extensively discussed in this paper (Morin and Conant 1990; see also Chapter II).

16 At the end of the 1986 field season, 10 ~. variabilis clumps containing nests with banded pairs that had produced young were semipermanently marked; 11 such bunchgrass clumps were marked in 1987.

These old nest sites were later mapped.

Located nests were usually checked daily until the clutch was complete, again on the 10th and 15th day after the first egg was laid, and thereafter daily until the clutch had hatched. Eggs were uniquely numbered within 24 hours after being laid, with a felt tip pen or finger nail polish. If the nest was located after eggs were laid, it was checked daily, or every other day, until the eggs hatched. Eggs in these nests were marked in order to track individuals, but were excluded from analyses requiring known laying order and clutch size. Egg lengths and widths were measured to 0.1 rom with calipers, and eggs were weighed to 0.1 g with a 10-gram Pesola scale. Only first day egg weights were used in analyses. All except 11 eggs over the three years were measured during May and June.

Newly hatched chicks were marked on the dorsal down with felt tip pens of different colors so that individuals from known eggs could be identified whenever possible. If the female parent was unbanded, I always tried to capture, band, and measure her just after the last or next-to-last chick had hatched, when she was easy to capture on the nest. I always covered the chicks/eggs with a light cotton cloth or porous plastic lid while I quickly banded the female.

When the hatching order of chicks was not known, I did not include the nest in hatch order analyses. The first morning that I saw a newly hatched chick (if I had seen the egg the previous morning), I assigned 17 it age "day one", since the egg had hatched in the previous 24 hours.

New chicks were measured on the first, second, and/or third day of life.

Chicks were measured every other morning thereafter until approximately

11 to 14 days of age, by which time they had been leg banded with a unique color combination and a numbered USFWS aluminum band. Nest checks and egg or chick measurements were not made during heavy rain or high winds.

The following measurements were made on chicks: unflattened right wing chord measured to the nearest 1.0 rom with a metric ruler; right tarsal length, beak depth, beak width, beak length, and sternum length, all measured to the nearest 1.0 rom with calipers. Occasionally, on 1- or 2- day old chicks, some measurements were taken with dividers.

Weights were taken to 0.1 g with a 10- or 50-gram Pesola scale, depending on the size of the chick. These weights and measures are described in detail elsewhere (see Chapter III). In 1987, detailed observations of feather emergence and chick development were made on five chicks.

Fates of ~ and Chicks

Nests were checked every few days, even after I was no longer handling the chicks, to determine the fate of offspring and the age of fledging. The definitions of the 17 possible fates are summarized in

.t~ Table 1-2. Each egg/chick was assigned only one fate. However, fates are not necessarily mutually exclusive, but represent a "best guess" based on the evidence. For example, some chicks that were assigned the fate of DEAD IN NEST may have also starved, but I only assigned the

18 STARVED fate in cases where it was clearly the cause of death. Many of the EGG DISAPPEARED may have belonged in other fate categories, such as

EGG CRACKED/PECKED/HOLES, but if I did not see enough evidence, I did not assign such a fate.

All data were statistically analyzed using routines from the personal computer Statistical Analysis System, version 6.03 (SAS 1988).

RESULTS

Weather

Daily air temperatures ranged from a minimum of 13.3 to a maximum of 36.1 °C (Table 1-1). Recorded wind speeds ranged from 0 to 48 km per hour, although storms with considerably higher wind speeds beyond our measurement capability occurred. Daily precipitation ranged from 0 rom to a high of 152 rom. Based on our rainfall data, as well as observations of vegetation status, 1986 can be considered an unusually wet year, 1987 a drought year, and 1988 an intermediate, probably more representative year (Figure 1-3).

Although not included in this paper, qualitative observations of plant phenology and finch feeding patterns indicated reduced food abundance during the 1987 breeding season (Morin unpubl. data).

Pair Formation and Maintenance

Laysan Finches exhibit delayed plumage maturation; males do not usually reach full adult plumage until their third year (Banks and

Laybourne 1977). Females attain a somewhat more "male-like" plumage as they age; e.g., older females have fewer dark streaks on the head (in some cases almost none) than younger females. In hatch-year and second-

19 year birds, both sexes in each age class appear similar in many aspects of their plumage, although females are usually more heavily streaked and have less yellow on the breast. The yellow on young females is usually paler than that of males. Hatch-year and second-year birds can usually be correctly sexed in the hand (especially when other characters such as bill depth and wing length are examined); however, I observed a great deal of variation. All males seen in banded, breeding pairs on Laysan had adult (after-second-year or ASY) plumage. However, Conant (pers. corom.) has seen second-year males paired and apparently breeding in the introduced Laysan Finch population at Pearl and Hermes Reef.

Courtship and pair formation were noticed well before the first nests of the season were found. For all three years, 8 March 1986 was the earliest date I recorded male singing, and 26 March 1986 was the earliest recorded pair formation. The earliest nest building activity

(female carrying grass) occurred on 19 March 1986. The earliest nest with eggs that I found in all three years was 2 April 1986. However, in

1987 the first eggs were not found until over a month later, on 13 May, even though I began searching for nests on 7 April.

Monogamy was the only breeding system observed. Once a pair had formed and a nest was under construction, the mates were either seen only together or each alone (N = 44 different pairs with known nest sites). At least nine banded pairs were resighted in subsequent years, indicating there is at least some year-to-year mate fidelity. In at least 12 cases, one or both members of a pair renested within the same season. Only once were both members of a previously identified pair seen together at a renest within the same season. More typically, only

20 one member of a pair was ever seen at the first or second nest, or one member of the pair was unbanded, so that positive pair identification was impossible. Repairing within a season was never observed. However, three individual birds repaired between seasons; two had spent two consecutive seasons with the same partner, but in the third (1988) season were seen with a new partner. The first case involved a female; her previous mate was also resighted in 1988. There is a small chance that the resight was actually another male that had accidentally lost a colored leg band, since her old mate had been banded with only three

(rather than the usual four) leg bands. In the second instance, a male was seen with a new mate in 1988, but his previous mate was not seen in

1988. The third individual, a female, changed mates between 1987 and

1988; her 1987 mate was not seen in 1988. These latter two repairings may simply represent replacement after the death of the previous partner. Because banded finches were seldom seen more than 0.8 km from where they had been banded (unpubl. data), it is probable that these missing mates would have been seen if they were alive.

The male frequently regurgitated food to his mate during courtship and during the laying and incubation period. The females quivered their wings while soliciting feeding with vocalizations similar to fledgling

Laysan Finches begging for food. Both sexes were seen singing and performing wing and tail "flips" near each other. (The wings were rapidly spread in and out and the tail was rapidly flipped up and down.)

The role of this display in pair formation is not understood. During

April 1987 (pre-breeding), I observed that females appeared to be soliciting males with these displays, but were often ignored. Both

21 sexes appeared to guard their mates and vigorously chased intruders.

Several times males were actually seen fighting. Fighting was often preceded by a "face off" display, where the males crouched face-to-face on the sand, raised their dorsal feathers and rump, and made movements toward each other with open beaks. Once, a female chased one male away and then begged from the remaining male, who then attempted to mount her. During the pairing and nest building period, males sang loudly from atop vegetation, usually the bunchgrass Eragrostis variabilis, although other plants such as Scaevola or Pluchea shrubs were sometimes used.

Duration of the Breeding Season

The breeding season is long, and varies somewhat by year in duration and onset. Renesting appears to be common. In 1986, I arrived in late February and found the first fresh nest on 2 April, and viable eggs were seen from 9 April until at least 26 July. In 1987 I arrived on 7 April, but did not find a nest under construction until 7 May. No hatch-year birds were seen when I first arrived in 1986 and 1987. The first 1987 nest with eggs was seen on 13 May, and the first known-age egg was laid on 17 May (Figure 1-4). Eggs were seen until at least 13

July.

In 1988 I first observed fledged hatch-year birds on 16 May; this indicated that a batch of eggs had been laid at least 45 days earlier, or by 2 April. The last 1988 eggs were seen on 13 August.

Egg laying peaked for all three years combined (491 eggs with known laying dates) during mid to late May, and this appeared to be true for

22 each separate year (Figure 1-4). In 1986, a secondary peak occurred in

April. A secondary peak in April (or earlier) 1988 must have also occurred, because hatch-year birds were already present upon my arrival that year. In 1987, only one breeding pulse was detected during my three and one-half month field season. However, since I was not present on Laysan for any entire year, other laying peaks may have occurred unobserved.

Nest Situation and Description

Nests were constructed of grass (Eragrostis variabilis) stems, roots, and leaves, and sometimes included stems of the non-native weed

Cenchrus echinatus. Nests occurred almost exclusively in ~. variabilis;

73.7% of the 278 natural nests with eggs or chicks that I followed in

1986-1988 were in ~. variabilis alone, and another 25.5% had ~. variabilis as either a canopy or substrate, mixed with one or more other plant species. Only two nests did not have Eragrostis as a component of the nest substrate or canopy (Morin and Conant 1990). Since Eragrostis clumps are preferred as a nest substrate, almost all nests occurred in the Eragrostis or Ipomoea-Boerhavia-Sicyos vegetation associations where the bunchgrass primarily occurs. Nests were usually in close proximity to one another. Infrequently, more than one complete or partially constructed nest was found in a single Eragrostis clump, but never two simultaneously active nests. Detailed descriptions of nest dimensions,

locations, and substrates have been presented elsewhere (Morin and

Conant 1990, see Chapter II).

23 Within- and Between-Year Nest Site Fidelity

Twenty marked Eragrostis clumps from 1986 and 1987 that had supported successful nests (and both members of the pair had been banded) were relocated in 1988. Thirteen of these clumps were collapsed or in the process of collapsing and were basically unsuitable as nest sites. Collapsed clumps were mostly dead and had fallen into pits formed when the sand beneath the clumps was undermined by the burrowing procellarid seabirds and weakened by the effects of rain and wind. Only one of the remaining seven clumps (a 1987 nest site) contained an active nest in 1988, but the 1988 female was not the same as the 1987 female, and the 1988 male was never sighted.

In 1987, only one of the ten staked clumps from 1986 was reused as a nest site, but it was used by a different pair. Seven of the clumps did not contain any nests, and the remaining two had collapsed. None of the banded pairs nested in exactly the same staked clump as the previous year.

Based on data from all three years, six banded pairs (out of 44 different pairs where both members were banded and the nest sites were known) nested with the same mate for two consecutive years in nearby

Eragrostis clumps. One pair built 1986 and 1987 nests only 3 m apart.

More typically, only one member of a pair was positively identified at the nest (the mate was either not seen or was unbanded). Even for these individuals, the subsequent year's nest could often be found by looking near the previous year's nest.

Within the same year, any renest(s) tended to be near each other, or even in the same nest cup. In 1988 one banded pair had two

24 unsuccessful nests before their third nest produced two fledglings; the first two nests were 34.6 m apart and the final nest was 56.5 m from the second nest. In 1987, a banded female renested in her original nest cup after her first clutch of four failed, but I left Laysan before I could determine the fate of her second clutch of three eggs. In 1986 a banded female reused her original nest cup after she successfully fledged a chick, but her second brood failed when the chicks starved.

Nest Construction and Adult Behavior at the Nest

It is not clear who chooses the nest site, but prior to nesting, females were frequently seen running on the ground, with the male in close pursuit. A female would quiver her wings in front of a clump of

Eragrostis variabilis before climbing in and investigating it. She did this repeatedly with a series of clumps, giving the appearance of

searching for a nest site. Once selected, the nest site might be guarded by both sexes, but a larger territory was not defended. Nest

sites of different pairs were often in close proximity. Relative to their nest sites, pairs foraged both near and far away, but did not

defend foraging areas. Foraging occurred island-wide, although some

spots appeared to be especially popular foraging "commons", probably

mostly because of aspects of the vegetation. Other factors such as

proximity to researcher tents (i.e. food crumbs), water availability,

the distribution of seabird nests, and the density of finch nests

probably affected foraging patterns.

Only'female Laysan Finches were seen carrying nesting material to

the nest, although a female was sometimes followed by her mate during

25 the nest building. I have seen a nest built in as little as two or three days, but four to seven days is probably more representative.

Since I often located nests after construction was approximately half completed, the sample size for complete constructions was too small to permit a good estimate of the full construction time. The longest construction times I observed were at least 27 and 29 days. In these two cases, the nests appeared to be abandoned during early construction and activity resumed later. Since the pairs for both nests were unbanded, it is possible that a different pair finished the nest.

After a nest seemed completed, the pair frequently left the site for several days, presumably to forage. After such an absence, the female was often found sitting in the empty nest cup in an incubating position. When I found a female in a nest cup after such an absence, I almost always found her first egg the next morning. Generally, females appeared to begin daytime incubation as soon as the first egg was laid.

However, on one occasion I checked a nest the night (2200 hours) after the first egg was laid and found that the female was not on the nest.

However, she was asleep on the nest the next night at the same time, after she had laid her second egg. Based on three night observations of a nest from 1987, the female brooded the chicks at night until they were at least 17 days old. The male was not seen during these nocturnal visits.

Males do not have brood patches, and only females were seen incubating. During egg laying and incubation, the female seldom left the nest site and was fed by the male by regurgitation. When he came to feed her, she usually left the nest and was fed by him nearby (within 3

26 m). Both the male and female seemed uneasy if I observed such feeding, and positioned themselves so that my view was blocked by vegetation. If the female noticed me when she was returning to the nest after a feeding bout, she usually would move instead to an Eragrostis clump that did not contain her nest, and then to another, and another, etc. I sometimes found nests by going to the clump that the female seemed to avoid.

As the eggs began to hatch, the female was especially reluctant to leave the nest. However, once hatching was completed, the female was much less "tight" on the nest.

When nesting was at its peak, females were seen less frequently because of their incubation schedules. At this time, adult males were unusually abundant in camp and were easily caught in food-baited traps and at water stations during dry conditions. Some of the foraging behavior for food and water displayed by the males occurred at a frenzied pace after eggs hatched and the chicks were growing. The female did not substantially help with feeding the chicks until they were several days old. Even then, the male seemed to do most of the feeding. Both the male and female fed the young by regurgitation.

Fecal sacs were removed by both parents, and were sometimes seen being carried away from the nest. This behavior persisted until the chicks were at least 16 days old, decreasing as the chicks aged. By the time the chicks had fledged, there was often feces all around the rim of the nest, which had become flattened due to the activity of the chicks.

27 ~ Characteristics and Laying Interval

Laysan Finch eggs resembled other Hawaiian honeycreeper eggs in coloration and spotting; they were a light cream color with maroonish­ brown speckles, often more heavily concentrated on the wide end of the egg (Berger 1972). One egg was laid daily, usually within a few hours of sunrise, until the clutch was complete. Infrequently, a day was skipped during the egg laying process.

The average egg length was 2.21 cm (N=568; 8.0.=0.109) and average egg width (or breadth) was 1.65 em (N=568; 8.0.=0.064). Based on the

Hoyt (1979) equation for egg volume estimation where:

Volume = 0.51 (length) (breadth) (breadth), the average Laysan Finch egg volume was 3.07 ± 0.06 cc. The average weight for an egg less than 24 hours old was 3.16 g (N=449, 8.0.=0.330); only fresh egg weights were used in analyses. The smallest and largest fresh egg weights that yielded viable fledglings were 2.4 g (N = 2) and

3.7 g (N = 2); the former weight was only 64.9% of the latter.

The distributions of fresh egg weights for all three years pooled, and for each year taken separately, were similar (Figure 1-5). The overall skewness was -1.195 and the overall kurtosis was 5.515 (N=449).

The only year during which I have egg measurements from April was

1986. There were no April eggs in 1987 and I did not arrive on Laysan until May in 1988. The mean fresh egg weight per nest from the April

1986 eggs was 3.18 g (N=23 nests, S.D. =.301) and the mean fresh egg weight per nest from the May and June 1986 eggs was 3.16 g (N=17 nests,

8.0. = .296). The overall distribution of mean fresh egg weights per nest, as well as the 1986 distribution of mean fresh egg weights per

28 nest, was not normal until after the April 1986 eggs were removed. A

Mann-Whitney test indicated that there was no significant difference between the two 1986 groups (Z=-.4254, P = .671). Therefore, I continued to combine the 1986 eggs.

The mean fresh egg weight per clutch was averaged for each of the three years; these means differed significantly among the three years of data. The 1986 mean was 3.17 g (N =40 nests, S.D. = .295, mode 3.2),

the 1987 mean was 3.10 g (N =51 nests, S.D. = .245, mode 3.1), and the

1988 mean was 3.24 g (N = 22 nests, S.D. = .281, mode 3.1). A Kruskal­

Wallis test performed on these data yielded a Chi square of 6.43, P =

.04. A Tukey test indicated that the mean egg weights per clutch for

1987 and 1988 were significantly different at the 0.05 level.

There were different trends for mean fresh egg weight among the

clutch sizes in different years (Figure 1-6). In 1986, there was a

trend of increasing mean egg weight with increasing clutch size. A

Mann-Whitney test indicated that the mean egg from a four-egg clutch in

1986 was significantly heavier than the mean egg from a three-egg clutch

(Z=2.273, P=0.023). In the drought year of 1987, the trend was

reversed. In that year, eggs from three-egg clutches were significantly

bigger than those from four-egg clutches (Z=-2.776, P=.006). In 1988,

no significant difference could be detected between the eggs weights

from three- and four-egg clutches, probably due primarily to the smaller

sample sizes (Z=-.597, P=.550).

A two factor unbalanced ANOVA was performed on ranked fresh egg

weights from three- and four-egg clutches only. The early eggs from

1986 were omitted in order to normalize the data. Year, clutch size,

29 and the interaction of year and clutch size were specified as the model effects. Type III partial sums of squares indicated that only the effect of year, and the interaction of year with clutch size, accounted for significant effects on egg weights (F = 5.21, P =.006 and F = 5.63,

P = .0002, respectively).

The body weight of a laying female (N=26) was not correlated with the mean fresh egg weight of her clutch (Pearson Correlation Coefficient

=0.079, P=0.703) nor with the total egg weight for her clutch (Pearson

Correlation Coefficient = 0.137, P=0.505).

During the study, several unusual eggs were seen. In 1987, three nests each had at least one defective egg. These eggs appeared to have unusually thin shells, through which I could see the yolk and/or albumin moving inside with an oversized air space. None of these eggs hatched.

A very tiny egg weighing only 0.9 g was laid in 1987. This egg and the other egg in the clutch (also unusually small at 2.0 g) did not hatch and were salvaged.

Clutch Size and Broods per Year

For clutches of known size, three eggs was the mode for all three years (total N=166; see Figure 1-7), and 3.19 eggs per clutch was the overall mean (S.D.=0.696). Although the result was close to significance, a Kruskal-Wallis test could not detect any difference among the three years in the mean clutch sizes (Chi square = 5.531, P

=0.063). In 1986 the mean clutch was 3.17 eggs (N=75, S.0.=0.742, range

1 to 5), in 1987 the mean clutch was 3.08 eggs (N=61, S.0.=0.557, range

30 2 to 4), and in 1988 the mean clutch was 3.47 eggs (N=30, S.D.=0.776, range 2 to 5).

When the single one-egg clutch was excluded, and the 6 five-egg clutches were combined with the four-egg clutches, a contingency table indicated that there was no association of clutch size with year (N=165 nests, Chi Sq.= 5.91, P=0.206).

For the 43 nests (all three years combined) with complete nest data and female parent measurements, clutch size varied with the amount of female furcular fat (G-test, G(adj.)=4.019, P < .05). Females with a furcular fat status of one or two had more clutches of two or three eggs than clutches of four or more eggs (15 vs. 6), whereas females with a furcular fat status of three or four had fewer clutches of two or three eggs and more clutches of four or five eggs (9 vs. 13). Females with a furcular fat status of two or lower composed 31% (5 of 16) of this sample in 1986, 65% (13 of 20) of this sample in 1987, and 43% (3 of 7) of this sample in 1988.

The mean female body weight for 1986 was 32.5 g (n=16), for 1987 it was 34.0 g (n=19), and for 1988 it was 33.5 g (n=6). These weights were not significantly different among the three years (ANOVA, P=.438)i however, the 1987 females had a large standard deviation (4.35 vs. 2.31

in 1986 and 1.34 in 1988), indicating much variability among their weights.

Based on the average egg weight (3.16 g) and average clutch size

(3.19 eggs), the estimated average clutch weight is 10.08 g. This is approximately 30.8% of the mean adult female body weight of 32.7 g

(based on N=120 females three years old or older). The maximum clutch 31 size of five, with an estimated clutch weight of 15.8 g, corresponds to only 48.3% of "'"the mean adult female body weight. This production is much less than the 95% to 110% of female body weight reported for clutch

weights in 10 species of fringillids (Amadon 1943; Rahn et ale 1975).

Finches are capable of raising at least two broods a year, although

my study cannot reveal what proportion of the pairs actually did so. Of

the 12 banded birds (eight individuals and two pairs) which I knew had

at least two clutches within a single year, four individuals and one

pair fledged young from their first clutches and the remaining four

individuals and the other pair did not. Four individuals and one pair

fledged young from their second clutches, while two individuals and the

other pair did not, and two had unknown nest outcomes. One of the pairs

failed in both their first and second nest attempt but fledged young

from their third nest. Only four of the birds (two individuals and one

pair) were known to have fledged at least one chick from both their

first and second clutches.

Duration of Incubation

The mean incubation period from the time an egg was laid until it

hatched was 15.69 days (N=191, 8.D.=0.662). The modal incubation period

was 16 days (Figure 1-8). I generally checked a hatching clutch only

once a day, usually in the morning. All three of the eggs with 14-day

incubation periods were laid after the female had apparently skipped a

day of laying; i.e., I did not see the egg on the morning when it

"should" have been laid but did see it the next morning. Either these

eggs were actually laid later in the day on which they "should" have

32

------_.- .. been laid (in which case they hatched after 15 days of incubation), or some mechanism such as prehatch clicking (Orent 1975) stimulated early hatching. I had accidentally dropped one of the two eggs with 1S-day incubation periods, which may have somehow affected its development.

The mean incubation period shortened from the first laid egg to the last laid egg: the average length of incubation for the first egg was

16.16 days (N=55; 5.0.=0.631), for the second egg 15.63 days (N=59;

5.0.=0.522), for the third egg 15.40 days (N=57; 5.0.=0.593), for the fourth egg 15.32 days (N=19; 5.0.=0.582), and for the single fifth egg on which I had complete data, the incubation period was 16 days. These incubation periods are significantly different when analyzed with a

Kruskal-Wa11is test (Chi square = 45.40, P = .0001).

Of 18 eggs from multiple egg clutches where only one egg in the clutch hatched (the hatched eggs ranged from the first to the fifth in , laying order), the average incubation period was 15.67 days (5.0. =

0.767). This is very similar to the overall mean incubation period.

A prediction of incubation period can be made using Rahn and Ar's

(1974) equation:

Log I = Log 12.03 + (.217) (Log W) where I = incubation period in days, W = weight of a fresh egg in grams, and logarithms are in base 10. For Laysan Finch eggs with an overall average fresh weight of 3.16 g, the predicted period is 15.43 incubation days. This predicted period is very similar to the mean observed period from my data.

33 Hatchability, Hatching Success, and Hatching Order

For all three years combined, the hatchability of eggs from known­ size clutches was 40.6% (i.e. 40.6% of all eggs laid, hatched).

Hatchability was similar during each of the three years (Table 1-3).

The low hatchability is due to the large number of eggs that disappeared prior to their anticipated hatch date.

A traditional measure of nest success is the percent of nests that hatch at least one egg (Hensler 1985). My data indicate an overall nest success rate of 56.4% for the three years combined (N=166).

However, nest success varied considerably among time periods within a year. For clutches laid after April in 1986, nest success was 62.5%

(N=48). The data from 1987 and 1988, which were gathered mainly in May and June, indicate similar nest success rates of 62.3% (N=61) and 63.3%

(N=30), respectively. For nests from April 1986, the success rate was only 22.2% (N=27); success was probably affected by the adverse weather conditions in April and early May (Figure 1-3).

The percent of nests in which all eggs hatched was low, primarily because of partial clutch losses prior to hatching rather than egg infertility. For all three years combined, every egg hatched in only

18.7% of the nests (31 of 166).

The eggs of Laysan Finches hatch asynchronously, although there is some variation in the duration of the total hatching interval. In a few nests, eggs hatched almost synchronously (within 24 hours), but usually the hatching of a clutch was spread over 2 or more days, with eggs hatching in the order laid (see Chapter III). Only one egg out of 151

(with both laying and hatching order known) hatched out of sequence.

34 This egg was laid first but hatched third in a clutch of three. It was also the smallest of the three in volume and weight. However, in other clutches with similar size distributions, the eggs hatched in the laying order.

Nestlings, Fledglings, and Post-fledging Dependent Period

Newly hatched chicks are easily identified because of their small size and wet, matted down, which dries within half a day. At hatching there is light grey down on the capital, spinal, femoral, alar, humeral, and part of the ventral feather tracts. There is no obvious down on the crural or caudal tracts. Chicks have yellow flanges and bill, and the inside of the gape is lavender with patches of red on the palate.

Chicks develop rather slowly compared to temperate passerines; pin feathers do not begin to emerge until 4 days of age. Feathers do not begin to unsheathe in most of the feather tracts until 10 days of age, and even later for the crural, capital, and caudal tracts (Table 1-4).

Because of the asynchronous hatching order, there is usually a size hierarchy among the nestlings within a clutch. The older chicks are almost always larger simply because they have received more feedings.

The size differences appear to be maintained until the chicks approach fledging age. The older chicks are probably better able to compete for parental food because of their longer legs and larger overall size.

A chronology of other characteristics in the development of nestlings up to 16 days old is summarized in Table 1-5. Chicks abandoned the nest when handled after 14 days of age, so I seldom banded chicks in the nest after this age. The few times that I banded only

35 some of the chicks in a brood because one or two escaped, I usually saw the banded chicks later in the company of the "correct" number of unbanded chicks, so it appears that clutches can regroup after such a disturbance.

Chicks gradually left the nest cup and perched on the nest rim as they aged; older chicks did so earlier than younger chicks.

Infrequently, after a disturbance, chicks moved from the nest clump to a nearby clump before the normal fledging age. Fledging was a gradual process; my definition of a fledged chick is one that no longer associates mainly with the nest, even though it may still frequent the nest clump of Eragrostis between brief forays outside. Chicks usually fledged between 22 and 26 days of age (N=37). A few unusually small chicks took longer than normal to fledge (e.g. one 1987 chick took 29 to

33 days). For about a week to ten days after fledging, chicks stayed close to an Eragrostis clump or a bush (e.g. Scaevola or Pluchea), either the nest clump or a clump or bush within 10 m of the nest clump.

During this period, the parents left the fledglings in the clump or bush, and returned periodically to feed them, calling them out of hiding with contact calls. Fledglings at this stage had a distinctive appearance. They retained the obvious yellow bill flanges, and had a peculiar, fluttery, bat-like flight. Gradually, the fledglings were seen more and more frequently, openly following and begging from one of the parents. The male parents were followed as often as the females, or more so. The parents sometimes split the brood, possibly because at this stage the parents often forage separately.

36 After fledging, the chicks depended on their parents for food for at least another three weeks. The oldest known-age fledgling I saw being fed by a parent was 45 days old. It was not uncommon to observe fledglings more than 40 days old begging from adults, although the likelihood of parental feeding dropped off as the chicks aged.

Occasionally fledglings were seen begging from adults other than their parents, but only once did I see a banded fledgling fed by an adult other than a parent. Fledglings begged with a loud, persistent call that is a cross between "chert" and "chonk", and flapped one or both wings. As they followed a parent, begging, they also observed the parent foraging. I often saw fledglings watch and then sample the same plant or plant part that the parent had just eaten. It appeared that chicks learned what to eat by following and watching adults forage; however, they probably also learned by trial and error. I observed several hatch-year birds trying to crack small bits of plastic marine debris; I even observed a few fights over especially attractive pieces of plastic. Small, perfectly round pieces seemed to be favored. By the time fledglings were 40 or more days old, they were regularly seen alone feeding themselves, or in the company of hatch-year birds other than their siblings. By about three months of age, hatch-year birds have been seen to move 0.8 km or more within a single day. Based on observations of known-age, recaptured fledglings, the yellow bill flanges are no longer obvious at a distance by about two and one-half months of age.

37 ~ and Chick Mortality

Overall, 69.3% of all eggs that hatched survived to fledge.

Hatchling survival was similar during each of the three years (Table 1-

3) •

The mean numbers of hatchlings per nest and fledglings per nest for each clutch size for the four breeding peaks are shown in Figure 1-9. I have separated 1986 into early and late seasons, since weather during the early season (Figure 1-3) had a devastating effect during that period, regardless of clutch size. In contrast, the late 1986 season

showed an increase in hatchlings and fledglings per nest with an

increase in clutch size. During the drought year of 1987, all the two­ egg clutches failed, and the four-egg clutches did no better than the three-egg clutches. Except for the two five-egg clutches, 1988 showed moderate increases in the mean hatchlings and fledglings per nest with

increasing clutch size.

Female parents with a fat status of one or two tended to have fewer fledglings per nest (mean=1.32, N=19 nests), whereas females with a fat status of three or four tended to have more (mean=2.10, N=20 nests). A t-test showed that the difference between these two fledging rates was at the borderline of significance (t=2.018, where P=O.OS corresponds to t=2. 026) • However, females with lower fat status had fewer eggs per clutch than females with higher fat status.

I handled a total of 306 eggs or chicks from clutches of unknown size and an additional 530 eggs from clutches of known size (Tables 1-6,

1-7, and 1-8). Each egg or chick was assigned one of seventeen possible

"fates" (Table 1-2) based on my brief but frequent visits to the nests.

38 The two most common fates for known and unknown-size clutches in all three years combined were EGG DISAPPEARED (24.9% or 208) and CHICK

FLEDGED (34.0% or 284). Four of the 284 fledged chicks are known to have died later in the same field season. Most of the eggs in the EGG

DISAPPEARED category probably belong in the more specific categories of

EGG CRACKED/PECKED/HOLE and DIED PIPPING, but could not be assigned to these categories due to lack of evidence.

Because of underestimation bias associated with nests found after a clutch is completed or found during the chick stage (Mayfield 1975,

Hensler 1985), I have eliminated such problem nests from the analyses which follow, unless otherwise stated. I have only included nests with known-size clutches, where I marked all eggs as they were laid. The exceptions are a very few clutches of five eggs. I included these as known-size clutches because, based on the many nests I have observed, five appears to be the maximum clutch size. I also kept track separately of the fates of eggs and chicks from nests of unknown clutch size, where at least one, and as many as four eggs were laid (Table 1-

6).

The overall fates of eggs from clutches of known size were primarily EGG DISAPPEARED 33.0 % (N=175), CHICK FLEDGED 28.1 % (N=149), and EGG ADDLED 10.2 % (N=54). The other 14 fates each accounted for less than 4% of the total 530 eggs from known-size clutches.

In each of the three years, egg fates from known-size clutches were mostly EGG DISAPPEARED and CHICK FLEDGED; together these fates accounted for three-fifths of each year's total (Table 1-7). However, of the 83 eggs from known-size clutches laid in April 1986, 47 (56.6%) 39 disappeared, 35 (42.2%) suffered various other fates, and only 1 (1.2%)

fledged a chick. This subset of the 1986 eggs accounts for 13 of the 18

EGG ABANDONED, 6 of the 10 CHICK STARVED, and 6 of the 10 EGG

CRACKED/PECKED/HOLES for that year (Table 1-7).

When the egg fates from known-size clutches were combined for all

three years (N=530), there was a tendency for the percent of eggs with the fate EGG DISAPPEARED to decrease as the clutch size increased (Table

1-8). In contrast, there was a tendency for the percent of eggs with

the fate CHICK FLEDGED to increase as the clutch size increased. I omitted the single clutch of one egg, and analyzed success (fledged

versus nonfledged eggs) for the other four clutch sizes. A two by four

contingency table showed that fledging success was not independent of

clutch size (G = 14.702, P = .002). However, analysis of only the

three- and four-egg clutches, and only the three-, four-, and five-egg

clutches, indicated that fledging success was independent of clutch size

(G = 2.266, P = .132; and G = 2.272, P = .321, respectively). Two-egg

clutches had significantly lower fledging success.

In 1986, the percent of eggs producing fledglings increased as the

size of the clutch increased (Figures 1-9 and 1-10). This was the only

year when an egg's prospect for fledging success was associated with

clutch size for three- and four-egg clutches (Chi square = 44.67, P <

.001); a higher percent of eggs from four-egg clutches produced

fledglings. Interestingly, in 1987 I found no five-egg clutches, and

none of the eggs in two-egg clutches produced fledglings. In the 1987

three-egg clutches, 33.3% of the eggs produced fledglings, whereas only

29.2% of the four-egg clutches did so (Figure 1-10). In 1988, four-egg

40 clutches had the highest percent of eggs that produced fledglings (35.0% vs. 24.4% for eggs from three-egg clutches).

Of the 165 known-size clutches from all three years, 149 fledglings were produced, yielding an overall average of 0.90 fledglings per nest.

Of these 149 banded fledglings, only five have been seen again a year or more later. All five of these birds were from four-egg clutches. An additional five individuals from the total 135 banded fledglings from unknown-size clutches have also been seen a year or more later, making a total of ten resighted fledglings. Therefore, survival for the first year was at least 3.5% (10 out of 284).

DISCUSSION

Weather and Reproduction

Laysan Island has a harsh environment and a fairly simple ecosystem. There are wide fluctuations in the temperature and moisture, as well as wind direction and intensity. These climatic factors alter the vegetation indirectly by regulating plant growth, or directly by means such as burying vegetation under sand during wind storms or flooding vegetation when the interior hypersaline lake receives large amounts of rain.

Laysan Finches are omnivores and eat some part of almost every plant on the island, as well as invertebrates, carrion, and seabird eggs. Finch feeding probably has a significant effect on some of the plant species and other animals, especially when finch populations are high (pers. obs.). However, their catholic diet does not protect the finches from some effects of weather. Adverse weather may cause almost

41 total nest failure. In April and May of 1986, Laysan received over 218 rom and 241 rom of rain, respectively (Figure 1-3). The rain in mid-May

raised the level of the lake at least 0.3 m and caused extensive

flooding, which extended into three of the plant associations where

finches regularly foraged (Pluchea and sesuvium-Heliotropium-Cyperus) or

nested (Ipomoea-Soerhavia-Sicyos). Several early nests were completely

destroyed by flooding, but even more destructive were the winds that

accompanied these storms. Many of the early nests had normal chicks that apparently starved to death (i.e., stopped growing or actually lost weight and had no body fat; Morin unpubl. data). Not only did rain and wind during these storms appear to stress the young chicks thermally, but the parents seemed unable to feed them adequately. Probably the wind and rain prevented the parents from foraging efficiently, especially since their energy requirements, as well as those of the chicks, probably rise during such conditions. From the 83 eggs hatched

in known-size clutches in April 1986, only one chick fledged. More than half of the eggs disappeared. Most of the other deaths occurred as abandoned eggs, eggs with holes (e.g. pecked eggs), or starved chicks.

An unusually high proportion of this confirmed mortality for the entire

1986 breeding season occurred during that short time interval of severe weather (Table 1-7).

Laysan experienced a dry year in 1987 (Figure 1-3), and weather once again influenced finch reproduction. The level of the hypersaline

lake was low; many sections of the lake bed were exposed and dry enough to walk across. Various observations of vegetation as well as total reproductive failure by the Laysan Duck (Anas laysanensis), further

42 confirm that 1987 was a drought year (pers. obs., A. Marshall pers. comm.). The native cucurbit vine (Sicyos maximowiczii), an important food source for the finch in the Ipomoea-Boerhavia-Sicyos vegetation association, decreased in abundance relative to 1986, and the morning glory, Ipomoea pes-caprae, was more abundant. Another important finch food, Portulaca spp. (probably ~ oleracea or a hybrid), which was abundant in 1986, was absent in 1987 until after a 13 June rain, when it began sprouting. In addition, many of the areas that were flooded the previous year had not yet recovered their vegetation, so that the potential finch foraging area was reduced not only in quality but also quantity. Of the three years studied, 1987 was the only one in which the finches did not attempt breeding early in the season. It also had the highest percent of addled and defective (e.g. malformed) eggs, eggs that died during pipping, and chicks that died in the nest from unknown causes (Table 1-7). Since relative humidity affects water loss from eggs, thus affecting hatchability (Drent 1975), the addled eggs and

chicks that died pipping may have been partially a result of the dry

conditions, as well as nutritional limitations.

Seven of the eight defective eggs that I identified in all three

years occurred in 1987. The defects of these eggs (unusually small size

or malformed shell and contents) suggest nutritional deficiencies. The

mean weight of eggs in 1987 was also significantly lower than in the

other years (3.10 g versus 3.17 g and 3.24 g). Usually females under

nutritional constraints lay relatively smaller eggs (O'Connor 1984).

Ewald and Rohwer (1982) found that experimentally providing food to wild

passerines allowed them to lay larger eggs, and Briskie and Sealy (1990)

43 found that eggs had a greater volume in years of naturally more abundant food.

The mean clutch size in 1987 was also the smallest of the three years (3.08 eggs versus 3.18 and 3.47), although the difference was not significant. No five-egg clutches were found in 1987, although they were found in both 1986 and 1988. This reduction in clutch size may have also been a response to nutritional stress caused by the drought.

In 1987, 65.0 % of the female parents of known-size clutches where

female fat status was known had a fat status of two or less. If fat

indicates female condition, as is commonly believed (Murton and westwood

1977), and as egg weight data from this study suggest, then 1987 was the year when females were in worst condition.

The low population estimate of 5,201 ± 1,211 birds made in May 1988

(Morin 1988, see Chapter IV), further supports the idea that 1987 was a poor year for reproduction and subsequent recruitment; May estimates for

1986 and 1987 were 10,333 ± 1,796 and 10,911 ± 1,769, respectively

(Morin 1986, Morin 1987).

Baldwin (1953) reported adverse effects on breeding of Hawaiian

honeycreepers ('Amakihi, ,Apapane, and 'I'iwi) on Hawaii Island during

stormy, wet periods, when hard rains caused nest destruction and

abandonment and fewer young were produced. The effect of adverse

weather on breeding has also been documented in the Large Cactus Finch

(Geospiza conirostris) in the Galapagos Islands (Grant and Grant 1989),

over both short and long intervals of time. Grant and Grant found that

the unpredictability of weather and its effect on environmental

conditions and food supply exerted selection pressures on the breeding

44 strategy of the Large Cactus Finch; this is probably also the case for

Laysan Finches. The variability of weather has a major influence on the breeding success of Laysan Finches every year. This in turn affects the population size, and this process may partially explain the wide fluctuations in the population that have been documented over the past two decades (Dennis et ale in press).

Laysan Finch Breeding strategy

Like the Large Cactus Finch (Grant and Grant 1989), the Laysan

Finch seems to have a breeding biology that has both temperate and tropical characteristics. Unlike temperate passerines, breeding condition is not regulated only by photoperiod, but is at least influenced, if not induced, by food supply. During the 1987 drought, for example, breeding did not begin until May, a full month or more later than in the other two years. Food availability as a proximate control over the onset of reproduction has also been observed in other fringillid finches (Newton 1973). Under optimal environmental conditions, it seems feasible that Laysan Finches could breed almost year round. Indeed, a captive female Laysan Finch held in the Honolulu

Zoo during 1989 had a brood patch and an attending male during November.

Laysan Finches have delayed plumage maturation (Banks and Laybourne

1977), and on Laysan the males do not breed until they are at least two years old (pers. obs.). Such delayed maturity generally occurs when adult annual survival exceeds about 60% (Murton and Westwood 1977). In general, adult male Laysan Finches may have higher survival rates than adult females (Conant unpubl. data). Survival on Laysan between

45 hatching year and second year appears to be low; only 3.5% of banded chicks that fledged were resighted a year or more later. The probability of resighting a banded bird is rather high, since dispersal of banded finches away from the banding sites is low (Morin unpubl. data). Therefore, it is more likely that banded fledglings died rather than dispersed to a distant area.

In an unpredictable environment, the probability of fledging young can be increased by spreading reproduction over a long breeding season with the potential for multiple broods and by having a long reproductive lifetime (Murton and Westwood 1977). Laysan Finches have the potential for multiple broods within a year, and also have a long reproductive lifetime.

Laysan Finch eggs are lighter in weight than would be expected.

Rahn et ale (1975) proposed an equation to relate egg weight to adult body weight:

Log (egg weight) = Log (a) + b Log (adult weight).

Weights are given in grams and logarithms are in base 10. They concluded that the exponent, b, was a constant 0.675 for all avian groups, but that the proportionality coefficient, a, varied among the orders and families. Using my mean egg weight estimate of 3.16 g, and a mean female Laysan Finch body weight of 32.7 g (for N=120 females three years old or older, banded on Laysan Island 1986 to 1988, unpubl. data),

I derive an estimate of a = 0.300. This is considerably less than the value of a = 0.413 which Rahn et al. (1975) compiled for 13 species of fringillids. The low average estimated clutch weight (30% of the mean adult female weight) may be adaptive in an unpredictable environment.

46 If the female does not deplete all her reserves on anyone nesting attempt, she can remain ready to renest quickly if necessary.

Laysan Finches are monogamous within any year, and display at least some year-to-year pair and site fidelity. These are reproductive features that can be expected in long-lived species when both parents are needed to raise the young successfully (e.g. for defense against predation: Oring 1982, Freed 1986), when previous experience increases reproductive success (O'Connor 1984), or when opportunities to acquire new mates are limited (Freed 1987). Year-to-year nest site fidelity focused on a relatively small area, but was not confined to a single

Eragrostis clump. This flexibility in specific site selection may be adaptive, since my limited data on individually marked bunchgrass clumps show that they have a rather high turnover rate. However, the finches' familiarity with a general area should increase their foraging efficiency, because they will know where food was located in the past.

Laysan Finches do not maintain a territory, but defend the nest and nest substrate (Eragrostis variabilis) itself and also guard their mate.

These behaviors decrease the likelihood of extra pair copulation and nest parasitism. Nest parasitism, where a female lays an egg in another finch's nest, was never documented. It seems extremely unlikely, since females seldom leave their nests once laying begins. Although it remains to be determined whether males engage in sneak copulations, none were observed. However, male parents are frequently absent from the nest site while foraging and have ample opportunity.

The lack of territoriality in Laysan Finches is probably related to the scattered temporal and spatial distribution of potential food, as

47 well as the distribution of the preferred nest substrate Eragrostis.

The absence of territorial defense may also save energy. Some of the best feeding areas contained few or no nest substrates; conversely, the second-best nesting area was nearly a monoculture of bunchgrass. The preferred breeding area appeared to be the viney Ipomoea-Boerhavia­

Sicyos association, which contained both nest substrates and a variety of important finch foods (see Chapter II). In addition, this association is below and well inside the berm of Laysan's sandy shoreline, which possibly confers some microhabitat protection from wind and blowing sand.

Mortality, Asynchronous Hatching, and optimal Clutch Size

The bias inherent in survival and mortality estimates based on nests found after laying has commenced, or nests found with chicks, has been much discussed in the literature. Briefly, such nests give an overestimate of survival and an underestimate of mortality for offspring. The Mayfield Method and various other modifications have been suggested to correct such nest data (Mayfield 1975, Hensler and

Nichols 1981, Bart and Robson 1982, Hensler 1985). In this study, I found many nests before laying began (my "known-size clutch" nests), so that I could individually identify eggs, the chicks they produced, and their ultimate fates. I also followed nests of unknown-size clutches, whose fates are summarized in Table 1-6. Comparing Table 1-6 with Table

1-8 (known-size clutches) clearly demonstrates how uncorrected data from unknown-size clutches inflate the fledging rate and seriously underestimate the number of eggs lost. Also, the number of nests of

48 unknown-size clutches is not representative of the true clutch size distribution. The uncorrected data of Table 1-6 suggest that there were relatively fewer three-egg clutches than there actually were (Table 1­

8), and that there were relatively more four- or five-egg clutches in the unknown-size clutches. These discrepancies may be explainable if larger numbers of nestlings (e.g. four chicks) made so much noise in the nest that I was more likely to find their nest.

Death before fledging is the most likely outcome for a Laysan Finch egg. Only one-quarter to one-third of all eggs from known-size clutches in each year fledged a chick. The most common fate for the rest of the eggs was to "disappear". Laysan has no mammalian or reptilian predators. Predation on fledglings by Great Frigatebirds (Fregata minor palmerstoni) is probable, and egg predation by migratory Bristle-thighed

Curlews (Numenius tahitiensis) and Ruddy Turnstones (Arenaria interpres) is possible, although I never witnessed it. walker (1961) reported seeing a curlew carrying a freshly killed Laysan Finch. Some Laysan

Finches have an interesting alarm call that sounds very similar to an alarm call made by Bristle-thighed Curlews. I have heard this call given especially by adult males when they were in my hand while being banded. However, on 4 June 1988, I saw a group of five or six finches

(both sexes) doing wing flips and alarm calls while following a Bristle­ thighed Curlew. At the time, I speculated that this behavior was a distraction to keep the curlew away from their eggs and chicks in nests, or possibly a warning for the already-fledged chicks. Predation by finches on finch eggs may also be important. It was not estimated in this study, although many of the eggs in the EGG DISAPPEARED category

49 may have been predated (Table 1-7). Laysan Finches have been reported to prey intraspecifically on eggs (Dill and Bryan 1912). On one occasion, I saw an unbanded finch (either a second-year male or a female) eating the contents of a finch egg while carrying it away from one of the study nests. This intraspecific predation may be a case of interference competition, whereby potential competitors are removed

(O'Connor 1984). Although the possibility of infanticide cannot be excluded, mate replacement following possible egg predation was never observed.

In principle, the striking asynchrony in the hatching of Laysan

Finch eggs seems consistent with the brood reduction model for asynchrony in an unpredictable environment. If future environmental conditions are not predictable at the time an egg is laid, brood reduction (the selective starvation of younger, smaller chicks) may occur during poor conditions. In this way, at least some of the chicks in a clutch may survive (Lack 1954).

However, my data do not consistently support the brood reduction model (see Chapter III), but suggest it may be applicable during some years. In 1986, eggs from bigger clutches actually had a higher rate of fledging than those from smaller clutches (Figures 1-9 and 1-10). My

1987 data may appear to support brood reduction, since a lower percent of eggs from four-egg clutches fledged chicks than eggs from three-egg clutches during that drought year. If brood reduction occurs only in some years, a drought year with limited food would be a likely time to expect it. However, in 1987 the earliest laid eggs in the laying sequence were less likely to fledge a chick than the later laid eggs;

50 the brood reduction model predicts that the later laid eggs would be less likely to fledge. In addition, during the drought year, egg weights were significantly lower in four-egg clutches than in three-egg clutches (Figure 1-6). However, the mean number of hatchlings per nest was the same for both clutch sizes (Figure 1-9), suggesting that egg size (and therefore indirectly brood size) may be adjusted by the female's ability to form viable eggs in relation to her nutritional status.

Several studies have shown that nestling survival is related to the size of the egg from which it hatched. Slagsvold et ale (1984) found that eggs that did not hatch were significantly smaller than those that did. This could explain why the 1987 four-egg clutches, which had significantly smaller egg weights than the 1987 three-egg clutches, did

so poorly (Figure 1-9). During 1988, a year that seemed intermediate between 1986 and 1987 in both food availability and rainfall (Figure 1­

3), four-egg clutches had a higher fledging rate than three-egg clutches

(Figure 1-9 and 1-10). The 1988 mean egg weight from four-egg clutches also tended to be heavier than the mean from three-egg clutches, even though the difference was not significant (Figure 1-6). These fledging trends are not surprising if the relationship of mean egg weights between three- and four-egg clutches in the different years are

reexamined (Figure 1-6).

Several other explanations have been proposed for hatching asynchrony (Clark and Wilson 1981, Mead and Morton 1985, Magrath 1988,

Murphy and Haukioja 1986). Clark and Wilson (1981) proposed that if the daily rate of predation on nestlings is less than the rate on eggs, then

51

------._-- incubation should start with the first egg in order to minimize the period of time when eggs are vulnerable to predation, and to allow at least some (older) nestlings to escape. Laysan Finches could fit this model of hatching asynchrony. Without more detailed knowledge on the high percent of eggs that disappear (many of which were probably preyed upon), it is hard to evaluate the model. Females sit on the nest even before laying the first egg, and their continuous presence on the nest may be a potent deterrent to both intra- and interspecific avian predation. Asynchronous hatching may be adaptive for a variety of reasons for various birds (Murphy and Haukioja 1986, Slagsvold 1990); the evidence is not clear as to which apply for Laysan Finches (see

Chapter III).

Why is three eggs the overall modal clutch size, when in two out of the three years (1986 and 1988) there tended to be a higher percent of eggs from four-egg clutches that produced fledglings? If 1987 was a poor year due to the drought, it could be that the optimal clutch size in a poor year (three eggs) is smaller than the optimal clutch size in good years (four or five eggs; Figures 1-9 and 1-10). In late 1986 and all of 1988, the average three-egg clutch fledged 1.29 and 0.73 chicks per nest, respectively. The average four-egg clutch in late 1986 fledged 2.27 chicks per nest, and in 1988 1.4 chicks per nest. In contrast, three-egg nests in 1987 fledged 1.0 chicks per nest, and four­ egg nests fledged only 1.2; the advantage of expending the extra energy for the fourth egg in 1987 appeared to be minimal. As discussed earlier, Laysan Finch females lay a much smaller total egg mass per clutch (less than 50% of female body weight) than the 100% of female

52 body weight expected (Amadon 1943, Rahn et al. 1975), so laying an extra egg may represent a relatively small investment. However, in good years, there is an advantage to laying a larger clutch, because more

fledglings per nest attempt will be produced (Figure 1-9).

A similar situation has been described by Boyce and Perrins (1987)

for Great Tits in a fluctuating environment. They suggested that in the

long run it is advantageous for such birds to lay clutches smaller than the largest possible clutch size. This strategy may be illustrated in the Laysan Finch by the fledging success in clutches of five. During a good breeding season (late 1986), the two five-egg clutches were the most productive, but during a representative year (1988), fledging

success in the two five-egg clutches was as poor as the success from two-egg clutches (Figure 1-9). Murton and Westwood (1977) believed that the optimal clutch size is always smaller than the largest possible clutch size, and that the most frequent clutch size is that which on the average gives the most survivors. This suggests that demanding environmental conditions on Laysan (e.g. droughts) are a fairly common occurrence. Although there may be selection against four- and five-egg

clutches in the poorest years, counter selection during the best years probably keeps the clutch size from stabilizing at three.

Although I have only anecdotal supporting observations, it seems

likely that high ambient temperatures, intense insolation and low wind

speed may sometimes cause extreme ground level heating of the sandy

Laysan substrate, such that bigger clutches with older nestlings may become seriously heat stressed. Several times I have seen chicks

panting in crowded nests, and upon handling them, found them to be

53 unusually warm. It seems likely that occasional heat stress also places an upper limit on clutch size, at least in some years. Although its natural environment is rather desert-like, the Laysan Finch has apparently never evolved special heat tolerance (Weathers and van Riper

1982), even though it has a reduced basal metabolic rate and evaporative water loss - adaptations associated with dry, hot environments.

Comparison with other Hawaiian Honeycreepers

Laysan Finches have been grouped in the tribe Psittirostrini with the other finch-billed Hawaiian honeycreepers. The endangered Nihoa

Finch (Telespiza ultima) is certainly the most closely related species in plumage and habit, but little is known about its breeding biology.

Recently it has been suggested that another endangered Hawaiian honeycreeper, the Palila (Loxioides bailleui), may be congeneric with the Laysan Finch (Johnson et al. 1989). When my data are compared to breeding biology studies on the Palila (van Riper 1978, Pletschet and

Kelly 1990), the two species seem similar in many aspects. They share similarly-timed, long breeding seasons apparently tied to food resources as well as photoperiod, courtship (and chick) feeding by regurgitation, mate and nest guarding rather than traditional territory defense, year­ to-year nest area fidelity, female-only incubation, similar incubation length (15.7 days for Laysan Finch and 16.6 for Palila), similar age at fledging (21 to 27 days for Palila and 22 to 26 for Laysan Finch), and an extended period during which the fledglings are dependent on their parents (about a month). They differ in modal clutch size (two eggs for

Palila and three for Laysan Finch), degree of nest sanitation (Laysan

54 Finches remove fecal sacs for a larger proportion of the nestling phase), and participation in nest construction. Only female Laysan

Finches have been seen constructing the nest, whereas both male and female Palila have been seen doing so (van Riper 1978). Other than the

Hawaii 'Akepa (L. Freed pers. corom.), the Laysan Finch is the only

Hawaiian honeycreeper for which the female alone is reported to construct the nest. Breeding studies on Palila, 'Anianiau, 'Amakihi,

'I'iwi, and 'Apapane concluded that both sexes participate in nest construction (van Riper 1978, Eddinger 1970). In this regard, only the

Hawaii 'Akepa and the Laysan Finch (and possibly the extinct Greater

Koa-finch, Rhodacant~is palmeri; Perkins 1903) are known to be similar to the cardueline finches. However, across avian taxa there is considerable plasticity in life history parameters, making life history weak evidence for phylogenetic classification.

In summary, the phylogenetic relationships among the Hawaiian honeycreepers are still unclear, as are many aspects of their biology.

Much remains to be learned about the many factors that have caused and are causing endangerment and extinction among the Hawaiian honeycreepers, and insular avian species in general. My research indicates that stochastic weather events and predation are probably the two major current limiting factors for the Laysan Finch, unlike most of the other Hawaiian honeycreepers, which are limited primarily by direct and indirect effects from past and present human activities.

55

------._------LITERATURE CITED

Amadon, D. 1943. Bird weights and egg weights. Auk 60: 221-234.

American ornithologists' Union. 1983. Checklist of North American

birds. 6th Edition. Allen Press, Inc., Lawrence, Kansas. 877 pp.

Bailey, A. M. 1956. Birds of Midway and Laysan Islands. Denver Mus.

Nat. Hist., Mus. Pict. 12. 130 pp.

Baldwin, P. H. 1953. Annual cycle, environment and evolution in the

Hawaiian honeycreepers, (Aves: Drepaniidae). Univ. Calif. Publ.

Zool. 52: 285-398.

Banks, R. C. and R. C. Laybourne. 1977. Plumage sequence and taxonomy

of Laysan and Nihoa Finches. Condor 79: 343-348.

Bart, J. and D. S. Robson. 1982. Estimating survivorship when the

subjects are visited periodically. Ecology 63(3): 1078-1090.

Berger, A. J. 1969. The breeding season of the Hawaii Amakihi. Occas.

Papers, B.P. Bishop Mus. 24:1-8.

Berger, A. J. 1970. The eggs and young of the Palila, an endangered

species. Condor 72:238-240.

Berger, A. J. 1972. Hawaiian birdlife. Univ. of Hawaii Press,

Honolulu. 270 pp.

Berger, A. J., C. R. Eddinger, and S. C. Frings. 1969. The nest and

eggs of the Anianiau. Auk 86:183-187.

Boyce, M. S. and C. M. Perrins. 1987. Optimizing Great Tit clutch size

in a fluctuating environment. Ecology 68(1):142-153.

Briskie, J. V. and S. G. Sealy. 1990. Variation in size and shape of

Least Flycatcher eggs. J. Field Ornith. 61(2): 180-191.

56 Butler, G. D., Jr. and R. L. Usinger. 1963. Insects and other

invertebrates from Laysan Island. Atoll Res. Bull. No. 98. 30 pp.

Collins, M. S. 1984. Observations on the nesting of the Hawaii 'Akepa.

'Elepaio 45(1): 1-2.

Conant, S. 1985. Trip report - expedition to Laysan Island 28 February

to 14 April 1985. Unpubl. report, Dept. of Gen. Sci., Univ. of

Hawaii, Honolulu. 5 pp.

Crossin, R. S. (June), 1966. Notes on the Laysan Finch (Psittirostra

cantans). Leeward Island Survey No. 13. Unpubl. Report, POBSP,

Smithsonian Institution, Washington, D.C. 10 pp.

Dennis, B., P. L. Munholland, and J. M. Scott. In press. Estimation of

growth and extinction parameters for endangered species. Ecol.

Monographs.

Dill, H. R. and W. A. Bryan. 1912. Report of an expedition to Laysan

Island in 1911. U.S. Dept. Agric. Bio1. Surv. Bull. 42. 30 pp.

Drent, R. 1975. Incubation. pp. 333-420. In Avian Biology. Vol. V.

D. S. Farner, J. R. King, and K. C. Parkes (eds.). Academic Press,

Inc., New York.

Eddinger, C. R. 1970. A study of the breeding behavior of four species

of Hawaiian honeycreepers (Drepanididae). Ph.D. dissertation, Univ.

of Hawaii, Honolulu. 212 pp.

Ely, C. A. and R. B. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll Research Bulletin, No.

171. 361 pp.

57 Ewald, P. W. and S. Rohwer. 1982. Effects of supplementary feeding on

timing of breeding, clutch-size and polygyny in Red-winged Blackbirds

Agelaius phoeniceus. J. Anim. Ecol. 51: 429-450.

Fisher, W. K. 1903. Notes on the birds peculiar to Laysan Island,

Hawaiian Group. Auk 20 (4): 384-397.

Freed, L. A. 1986. Territory takeover and sexually selected

infanticide in tropical house wrens. Behav. Ecol. Sociolbiol. 19:

197-206.

Freed, L. A. 1987. The long-term pair bond of tropical House Wrens:

advantage or constraint? Am. Nat. 130: 507-525.

Freed, L. A., T. M. Telecky, W. A. Tyler III, and M. A. Kjargaard.

1987. Nest-site variability in the 'Akepa and other cavity-nesting

forest birds on the island of Hawaii. 'Elepaio 47(8): 79-81.

Freed, L. A., S. Conant, and R. C. Fleischer. 1987. Evolutionary

ecology and radiation of Hawaiian passerine birds. Trends in Ecol.

and Evol. 2(7):196-203.

Freed, L. A. 1988. Demographic and behavioral observations of the

Hawai'i 'Akepa on Mauna Loa. 'Elepaio 48(5): 37-39.

Grant, B. R. and P. R. Grant. 1989. Evolutionary dynamics of a natural

population: the Large Cactus Finch of the Galapagos. Univ. of

chicago Press, Chicago. 350 pp.

Hensler, G. L. and J. D. Nichols. 1981. The Mayfield method of

estimating nesting success: a model, estimators and simulation

results. Wilson Bull. 93(1): 42-53.

Hensler, G. L. 1985. Estimation and comparison of functions of daily

nest survival probabilities using the Mayfield method. pp 289-301.

58 In Statistics in Ornithology 29. Lecture notes in statistics.

Morgan, B. J. T. and P. M. North, (eds.). Springer-Verlag.

Hoyt, D. F. 1979. Practical methods of estimating volume and fresh egg

weight of bird eggs. Auk 96: 73-77.

Johnson, N. K., J. A. Marten, and C. J. Ralph. 1989. Genetic evidence

for the origin and relationships of Hawaiian Honeycreepers (Aves:

Fringi11idae). Condor 91: 379-396.

Kern, M. D. and C. van Riper III. 1984. Altitudinal variations in

nests of the Hawaiian honeycreeper Hemignathus virens virens. Condor

86:443-454.

Kramer, R. J. 1959. Some notes and observations on Laysan Island fauna

April 28, 1959 to May 1, 1959. Unpub1. notes, Hawaii State Div. of

Forestry and Wildlife (formerly Dept. of Fish and Game), Honolulu.

8pp.

Lack, D. 1954. The natural regulation of animal numbers. Oxford

University Press, London.

Lamoureux, C. H. 1963. The flora and vegetation of Laysan Island.

Atoll Res. Bull. No. 97. 14 pp + figures.

Magrath, R. D. 1988. Hatching asynchrony in altricial birds: nest

failure and adult survival. Amer. Nat. 131 (6): 893-900.

Mayfield, H. F. 1975. Suggestions for calculating nest success.

Wilson Bull. 87(4):456-466.

Morin, M. P. 1986. Laysan Finch report 1986 field season Laysan

Island. Unpub1. Rep., U.S. Fish and Wildlife Service, Honolulu. 12

pp.

59 Morin, M. P. 1987. 1987 field season Laysan Finch report Laysan

Island. Unpubl. Rep., U.S. Fish and Wildlife service, Honolulu. 8

pp.

Morin, M. P. 1988. 1988 Laysan Finch report Laysan Island. Unpubl.

Rep., U.S. Fish and Wildlife Service, Honolulu. 9 pp.

Morin, M. P. In prep. Growth and mortality of Laysan Finch (Telespiza

cantans) nestlings in relation to hatching asynchrony.

Morin, M. P. In prep. Laysan Finch nest characteristics, nest spacing,

and reproductive success in two vegetation types.

Morin, M. P. In prep. Laysan Finch population estimates in relation to

the annual breeding cycle and other variables.

Morin, M. P. and S. Conant. 1990. Nest substrate variation between

native and introduced populations of Laysan Finches. Wilson Bull.

102:591-604.

Murphy, E. C. and E. Haukioja. 1986. Clutch size in nidicolous birds.

pp. 141-180. In Current ornithology. Vol. 4. R. F. Johnston,

(ed.). Plenum Press, New York.

Murton, R. K. and N. J. Westwood. 1977. Avian breeding cycles. Oxford

Univ. Press, Oxford. 594 pp.

Newman, A. L. 1988. Mapping and monitoring vegetation change on Laysan

Island. M.A. thesis, Univ. of Hawaii, Honolulu. 234 pp.

Newton, I. 1973. Finches. Taplinger Publishing Co., New York.

288 pp.

O'Connor, R. 1984. The growth and development of birds. John Wiley

and Sons, Ltd., New York. 315 pp.

60 Olson, S. L. and H. F. James. 1982. Prodromus of the fossil avifauna

of the Hawaiian Islands. Smithsonian Contributions to Zoology. No.

365. Smithsonian Institution Press. 59 pp.

Olson, S. L. and H. F. James. 1986. The holotype of the Laysan Finch

Telespiza cantans Wilson (Drepanidini). Bull. Brit. Ornithol. Club

106: 84-86.

Oring, L. W. 1982. Avian mating systems. pp. 1-92. In Avian

Biology. Vol. 6. D. S. Farner and J. R. King (eds.). Academic

Press, New York.

Perkins, R. C. L. 1903. Vertebrata (Aves). Pp. 368-465. In Fauna

Hawaiiensis. Vol. 1, part 4. D. Sharp (ed.). The Univ. Press,

Cambridge, England.

Pletschet, S. M. and J. F. Kelly. 1990. Breeding biology and nesting

success of Palila. Condor 92: 1012-1021.

Rahn, H. and A. Ar. 1974. The avian egg: incubation time and water

loss. Condor 76: 147-152.

Rahn, H., C. V. Paganelli, and A. Ar. 1975. Relation of avian egg

weight to body weight. Auk 92: 750-765.

Rothschild, (L.) W. 1893-1900. The avifauna of Laysan and the

neighboring islands: with a complete history to date of the birds of

the Hawaiian Possessions. R.H. Porter, London. 320 pp.

SAS Institute Inc. 1988. SAS/STAT User's Guide, Release 6.03 Edition.

Cary, NC. 1028 pp.

Sakai, H. F. and T. C. Johanos. 1983. The nest, egg, young, and

aspects of the life history of the endangered Hawaii creeper. West.

Birds 14:73-84.

61 Scott, J. M., J. L. Sincock, and A. J. Berger. 1980. Records of nests,

eggs, nestlings and cavity nesting of endemic passerine birds in

Hawaii. 'Elepaio 40: 163-168.

Scott, J. M., C. B. Kepler, C. van Riper III, and S. I. Fefer. 1988.

Conservation of Hawaii's vanishing avifauna. Bioscience 38(4):238-

253.

Sincock, J. L. and E. Kridler. 1977. The extinct and endangered

endemic birds of the Northwestern Hawaiian Islands. Unpubl. Ms.,

U.S. Fish and Wildlife Service, Honolulu. 111 pp.

Slagsvold, T. 1990. Fisher's sex ration theory may explain hatching

patterns in birds. Evolution 44(4): 1009-1017.

Slagsvold, T., J. Sandvik, G. Rofstad, O. Lorensen, and M. Husby. 1984.

On the adaptive value of intraclutch egg-size variation in birds.

Auk 101: 685-697. van Riper, C. III. 1978. The breeding biology of the Amakihi (Loxops

virens) and Palila (Psittirostra bailleui) on Mauna Kea, Hawaii.

Ph.D. dissertation, Univ. of Hawaii, Honolulu. 166 pp. van Riper, C. III. 1980. Observations on the breeding of the Palila

Psittirostra bailleui of Hawaii. Ibis 122: 462-475.

Walker, R. L. 1961. Excerpts from leeward island journal, Sept. 2nd to

12th, 1961. Unpubl. Report, Hawaii Div. of Forestry and Wildlife

(formerly Dept. of Fish and Game), Honolulu. 4 pp. + 9 attach.

Weathers, W. W. and C. van Riper, III. 1982. Temperature regulation in

two endangered Hawaiian honeycreepers: the Palila (Psittirostra

bailleui) and the Laysan Finch (Psittirostra cantans). Auk 99: 667­

674.

62 Wilson, S. B. 1890. On a new Finch from Midway Island, North Pacific.

Ibis 33: 339-341.

Woodside, D. H. (Sept.), 1961. A report on a trip to the Laysan Island

September 1961. Unpubl. report, Hawaii Div. of Forestry and Wildlife

(formerly Dept. of Fish and Game), Honolulu. 42 pp.

United States Fish and Wildlife Service. 1984. Endangered and

threatened wildlife and plants. Dept. of the Interior. 50 CFR 17.11

and 17.12. U.S. Government Printing Office: 1984 0-447-430.24 pp.

63 TABLE 1-1- Average daily maximum and minimum temperature (degrees centigrade) and average daily wind speed (km/hr) •

1986 1987 1988 MARCH Ave. max. temp. 27.5 (n=27) Ave. min. temp. 19.6 (n=27) Ave. wind speed 14.8 (n=25) 11.0 (n=21)

APRIL Ave. max. temp. 24.0 (n=22) 28.0 (n=28) Ave. min. temp. 19.4 (n=22) 20.7 (n=29) Ave. wind speed 15.1 (n=30) 14.5 (n=22) 11.2 (n=28)

MAY Ave. max. temp. 26.2 (n=15) 24.1 (n=31) 27.9 (n=29) Ave. min. temp. 21.6 (n=15) 19.8 (n=31) 22.3 (n=29) Ave. wind speed 14.4 (n=17) 12.4 (n=31) 9.9 (n=29)

JUNE Ave. max. temp. 28.4 (n=30) 26.5 (n=30) 31.7 (n=29) Ave. min. temp. 22.4 (n=30) 21.8 (n=30) 23.8 (n=29) Ave. wind speed 15.7 (n=30) 9.7 (n=30) 8.5 (n=29)

JULY Ave. max. temp. 29.9 (n=31) 29.6 (n=31) 32.0 (n=31) Ave. min. temp. 23.8 (n=31) 23.9 (n=31) 24.3 (n=31) Ave. wind speed 15.5 (n=31) 7.7 (n=31) 15.5 (n=30)

AUGUST Ave. max. temp. 30.6 (n=31) 32.5 (n=30) Ave. min. temp. 24.8 (n=31) 24.0 (n=30) Ave. wind speed 14.5 (n=2) 8.8 (n=31) 9.6 (n=30)

SEPTEMBER Ave. max. temp. 29.8 (n=12) Ave. min. temp. 23.6 (n=12) Ave. wind speed 10.1 (n=l1)

64 TABLE 1-2. Abbreviations and descriptions of egg fates.

FATE DESCRIPTION CHICK DEFECT/ Chick died of injury, defect, or disease before it DISEASE fledged.

CHICK DEAD IN Chick died in nest prior to fledging from unknown NEST causes.

CHICK DEAD Chick found dead outside nest prior to fledging from OUTSIDE NEST unknown causes.

CHICK STARVED Chick starved to death (i.e. did not thrive or was neglected and thin) before it fledged.

CHICK Chick was seen after hatching, but chick later DISAPPEARED disappeared without a trace.

CHICK FLEDGED Chick apparently fledged.

CHICK NEVER Egg apparently hatched (seen during pipping) but chick SEEN was never seen.

EGG DROWNED Egg drowned due to rain or flooding.

EGG CRACKED / Egg found cracked or with holes from predation or PECKED/HOLES mechanical damage prior to hatch "due" date.

EGG ADDLED/ Intact egg seen a day or more past its hatch "due" date INFERTILE and nest was not abandoned.

EGG ABANDONED Intact egg seen a day or more past the day when the nest was apparently abandoned.

RESEARCHER Egg did not hatch due to damage or disturbance by DAMAGE researcher.

EGG DISAPPEARED Egg apparently did not hatch and disappeared; shells or yolk mayor may not have been found.

NEST DESTROYED Nest destroyed by seabirds, other finches, weather, or other mechanical damage.

EGG DEFECTIVE Egg was misshapen or not of normal composition.

DIED PIPPING Egg appears to have died during pipping; these eggs are not considered to have hatched.

UNKNOWN Unknown if egg hatched, or if chick fledged.

65 TABLE 1-3. Hatchability, defined as (number of eggs hatched / number of eggs laid) X 100, and hatchling survival, defined as (number of fledglings / number of hatchlings) X 100, for Laysan Finch eggs from known-size clutches.

YEAR NO. OF EGGS NO. HATCHED NO. FLEDGED HATCHABILITY HATCHLING LAID SURVIVAL

1986 238 94 65 39.5% 69.1%

1987 188 80 56 42.6% 70.0%

1988 104 41 28 39.4% 68.3%

OVERALL 530 215 149 40.6% 69.3%

66 TABLE 1-4. Development of feather tracts in Laysan Finch nestlings.

AGE (DAYS) TRACT 1 2 3 4 5 6 7 8 9 10 11 12 13 14

SPINAL

ALAR

VENTRAL

CAUDAL

CAPITAL

HUMERAL

FEMORAL

CRURAL

••• PIN FEATHERS APPEAR --- PIN FEATHERS UNSHEATHING

67 TABLE 1-5. Developmental patterns in young Laysan Finches.

AGE (DAYS) DEVELOPMENTAL 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 CHARACTERISTIC

SHUT OPENING FULLY OPEN OPENING OF EYES

NOT ABLE TO GRASP ==~~~~~~~~~==GRASPING DEVELOPED GRASPING

NO COWERING RUN AWAY FEAR RESPONSE

UNABLE SOME STANDING STANDING WELL ABILITY TO STAND TO STAND

YELLOW TRANSITIONAL BLACKISH COLOR OF BEAK

NONE SEEN UP TO DAY 16 PREENING

NONE SEEN UP TO DAY 16 WING FLAPPING

68 TABLE 1-6. Number (N) and percent (%) of the most common fates for eggs from unknown-size clutches in each of three years.

FATE

CLUTCH SIZE YEAR EGG CHICK OTHER TOTAL DISAPPEARED FLEDGED

N % N % N % N %

AT LEAST 1 EGG 1986 0 (0.0) 0 (0.0) 0 (0.0) 0 (0.0) 1987 1 (25.0) 1 (25.0) 2 (50.0) 4 (100.0) 1988 0 (0.0) 7 (50.0) 7 (50.0) 14 (100.0)

AT LEAST 2 EGGS 1986 0 (0.0) 1 (50.0) 1 (50.0) 2 (100.0) 1987 4 (28.6) 0 (0.0) 10 (71.4) 14 (100.0) 1988 0 (0.0) 14 (53.9) 12 (46.2) 26 (100.0)

AT LEAST 3 EGGS 1986 13 (33.3) 9 (23.1) 17 (43.6) 39 (100.0) 1987 7 (21.2 ) 9 (27.3) 17 (51. 5) 33 (100.0) 1988 6 (9.1) 43 (65.2) 17 (25.8) 66 (100.0)

AT LEAST 4 EGGS 1986 1 (5.0) 10 (50.0) 9 (45.0) 20 (100.0) 1987 0 (0.0) 9 (45.0) 11 (55.0) 20 (100.0) 1988 1 (1. 5) 32 (47.1) 35 (51. 5) 68 (100.0)

TOTAL NUMBER EGGS 33 135 138 306

69 TABLE 1-7. Number (N) and percent (%) of fates for eggs from known-size clutches in each of three years.

FATE 1986 1987 1988

N % N % N %

Egg Disappeared 82 (34.5) 56 (29.8) 37 (35.6) Chick Fledged 65 (27.3) 56 (29.8) 28 (26.9) Egg Addled 15 (6.3) 29 (15.4) 10 (9.6) Egg Abandoned 18 (7.6) 1 (0.5) 2 (1.9) Egg Cracked/pecked/Holes 10 (4.2) 3 (1.6) 5 (4.8) Chick Starved 10 (4.2) 4 (2.1) a (0.0) Chick Dead in Nest 5 (2.1) 10 (5.3) 2 (1.9) Chick Defect/Disease 2 (0.8) 3 (1.6) 5 (4.8) Nest Destroyed 7 (2.9) 3 (1. 6) 5 (4.8) Chick Disappeared 8 (3.4) 5 (2.7) 3 (2.9) Chick Never Seen 2 (0.8) 0 (0.0) 1 (1.0) Egg Defective 1 (0.4) 7 (3.7) a (0.0) Chick Dead outside Nest 1 (0.4) 2 (1.1) 2 (1.9) Researcher Damage 4 (1. 7) 1 (0.5) 1 (1. 0) Egg Drowned 4 (1. 7) 0 (0.0) a (0.0) Died Pipping 1 (0.4) 6 (3.2) a (0.0) Unknown 3 (1. 3) 2 ( 1.1) 3 (2.9)

TOTAL EGGS 238 (100) 188 (100) 104 (100)

70 TABLE 1-8. Number (N) and percent (%) of the most common fates for eggs from all known-size clutches in 1986, 1987, and 1988.

CLUTCH SIZE

ONE TWO THREE FOUR FIVE FATE N % N % N % N % N %

Egg Disappeared 1 (100.0) 27 (75.0) 99 (32.7) 39 (24.4) 9 (30.0) Chick Fledged 0 (0.0) 2 (5.6) 84 (27.7) 54 (33.8) 9 (30.0) Egg Addled 0 (0.0) 2 (5.6) 26 (8.6) 26 (16.3) 0 (0.0) Egg Defective 0 (0.0) 2 (5.6) 3 (1.0) 3 (1. 9) 0 (0.0) Egg Abandoned 0 (0.0) 0 (0.0) 17 (5.6) 4 (2.5) 0 (0.0) Egg Cracked/ pecked/Holes 0 (0.0) 2 (5.6) 8 (2.6) 5 (3.1) 3 (10.0) Chick Disappeared 0 (0.0) 0 (0.0) 14 (4.6) 1 (0.6) 1 (3.3) Other 0 (0.0) 1 (2.8) 52 (17.2) 28 (17.5) 8 (26.5)

Total number 1 (100.0) 36(100.0) 303(100.0) 160(100.0) 30(100.0) of eggs (N)

Total number 1 18 101 40 6 of clutches

71 Figure 1-1 The Hawaiian Archipelago.

0 1800 1700W 160 30' 30 N N Kure o 200 kIn Midway 'Pearl and Hermes Reef

Lisionski • • Layson Gardner, Pinnacles French Frigate Shoals -.J • '" • Necker • Nih 00 KAUAI Niihou/. OAHU Kc u l c ... _MOLOKA I Lanai·~MAUI

0 Kahoolawe .. 20 20 HAWAII

0W 0 1800 170 160 Figure 1-2

Map of Laysan Island showing primary study area•

..:..

Ephemeral ::. lake ~

~ Primary

Lake and mudflat

N

~ Vegetated Area

o I

73 .-CD II 250 z ..,0 II 1986 z CJ .. 1987 200 ~ 1988 z -c ~ I.J.. a 150 .., N m en II C'l ~ z II W z J- W 100 :i!: N ::::i .- II ....J- z :i!: 50

MAR APR MAY JUN JUL AUG SEP

FIGURE 1-3

Minimum monthly rainfall (millimeters). N indicates the number of days with data. Where no N is shown, no data were collected.

74 15

>-« 1986 0 10 0:: W a.. en 5 "w 0 15 30 15 30 14 29 APRIL MAY JUNE 15

>--c 1987 0 10 0:: " W L- a.. a. -c en 5 c > "w .~ " L- 0 -c 15 30 15 30 14 29 APRIL MAY JUNE 15

>-« 1988 0 .q- 10 ..- 0:: W ~ a.. c ~ en 5 "w 0 15 30 15 30 14 29 APRIL MAY JUNE FIGURE 1-4 Temporal distribution of egg laying. See "Duration of the Breeding Secaon " under Results for latest egg laying dates in each season. 75 25% - c=J 1986 (N=189) .. 1987 (N=176) ~ 1988 (N=84) 20% -

(/) o "w 15%- u, o I- Z ~ 10%- 0::: w D..

5% -

~l ~ rI r ~ ~ J. ~ ~ 10 «) 0) 0 or- N t') ~ It) co r-, 00 00 .. . " ..00 ...... C\J C'II t'II C'II C'II C'II C'II t') t') t') t') t') t') t') t') t") t') V 1\ EGG WEIGHT IN GRAMS

FIGURE 1-5 Distribution of fresh egg weights during three years. N indicates number of eggs.

76 4.0 • ALL YEARS A 1986 o 1987 V 1988

3.5 I 1

I- 3.0 I C) w I I 3=

2.5

-or It) o (ONO o en 0 0 NO .... CO o N ...... -or NCO.... ", ...... ", .... It)OIt) II II II II II II I II II II II II II II II II zzzz zz z z zzzz zzzz 2.0 ...... ------TWO THREE FOUR FIVE CLUTCH SIZE

FIGURE 1-6

Mean fresh egg weights and standard deviations by year in relation to cl utch size. N indicates number of eggs.

77 100 [:=J 1986 (N=75) 90 .. 1987 (N=61) 80 ~ 1988 (N=30) V) w :c 70 o I­ :;) 60 ...J U 1J.. o I­ Z W U 0:::: W a..

1 2 3 4 5 CLUTCH SIZE

FIGURE 1-7

Distribution of clutch sizes by year. N indicates number of clutches.

78

------100 - 90 - 80 - Cf) 70 o - "w 60 u, - 0 c::: 50 w - []J :::!: 40 :::> - z 30 - 20 - 10 - 0 I 14 15 16 17 18 INCUBATION PERIOD (DAYS)

FIGURE 1-8

Incubation periods in days for 191 eggs from 1986 to 1988 combined.

79 Q::: 5 b. Clutch of 1(N=1) W 0 Clutch of 2(N=8) []J 4 ~ IJ Clutch of 3(N=20) ::::> 3 e Clutch of 4(N=7) Z 2 CIutch of 5(N=1) Z '" -c w 1 :::E a EGGS HATCHED FLEDGED Early 1986

Q::: w 5 "'- []J 4 ~ 0 Clutch of 2(N=1) ::::> 3 IJ Clutch of 3(N=24) Z 2 o Clutch of 4(N=11) z Clutch of 5(N=2) -cw 1 =8 '" ~ a EGGS HATCHED FLEDGED Late 1986

Q::: w 5 ED 4 ~ 0 Clutch of 2(N=7) ::::> 3 IJ Clutch of 3(N=41) Z 2 o Clutch of 4(N=12) Z -cw 1 ~ a EGGS HATCHED FLEDGED 1987

Q::: w 5 []J 4 ~ ::::> 3 0 Clutch of 2(N=2) z IJ Clutch of 3(N=15) 2 Clutch of 4(N=10) Z o -c 1 Clutch of 5(N=2) W '" ~ a EGGS HATCHED FLEDGED 1988 FIGURE 1-9 Mean number of hatchlings and fledglings per nest for each clutch size in each reproductive period. N indicates number of clutches.

80 .. Fledge CJ Disappeared ~ Other N co IX) r') t'Il It) v '" IX) It) 0 It) N=...... f'...... -..- v 0 'V v v .... 100

90

80

(/) 70 o "W 60 I.L. o 50 I-­ Z W U 40 0::: W n, 30

20

10 o 2 3 4 5 2 3 4 5 2 3 4 5 1986 1987 1988 CLUTCH SIZE FIGURE 1-10 Fates of eggs in known-size clutches by year. N indicates number of eggs.

81 CHAPTER II

LAYSAN FINCH NEST CHARACTERISTICS, NEST SPACING, AND

REPRODUCTIVE SUCCESS IN TWO VEGETATION TYPES

INTRODUCTION

The location, spacing, and composition of avian nests profoundly affect the microclimate in which the egg develops (Drent 1983), as well as the egg's subsequent success (Rendell and Robertson 1989). Numerous factors influence nest characteristics and site choice, e.g. prevailing winds (Ferguson and Siegfried 1989), the density of vegetation (Leonard and Picman 1987), the distribution of available nest sites (Hagan and

Walters 1990, Kerpez and Smith 1990), and intra- and interspecific competition (Rendell and Robertson 1989). Many recent studies have attempted to summarize specific habitat attributes that are correlated with nest sites selected by various species (Burger and Gochfeld 1988,

McAuliffe and Hendricks 1988, McCallum and Gehlbach 1988, Rendell and

Robertson 1989). My objectives were to describe the nest characteristics, nest spacing, and relative reproductive success for

Laysan Finch nests in the two major vegetation associations on Laysan

Island, so that nest site selection can be better understood for this

species.

The Laysan Finch (Telespiza cantans) is an endangered, endemic

Hawaiian honeycreeper that occurs naturally only on the uninhabited

island of Laysan in the Hawaiian Archipelago. This species defends the

nest site during breeding but does not defend a larger foraging territory. Nests can therefore be close together and birds can forage

near other nests. In a previous paper (Morin and Conant 1990), we

82 documented that on Laysan Island the Laysan Finch shows a striking dependence on a single species, the bunchgrass Eragrostis variabilis, as a nest substrate. The few recent researchers to comment on Laysan Finch nest sites and characteristics had noted the same dependence (Crossin

1966, Ely and Clapp 1973, Sincock and Kridler 1977). otherwise, the finches' nesting habits have been little studied.

The Laysan ecosystem was subjected to severe human-caused impacts in the early part of this century (Ely and Clapp 1973). Although the vegetation has recovered substantially, non-native plants and invertebrates are present on the island, and some are spreading (Newman

1988). These non-native species may exert a long-term effect of unknown magnitude on the finches and their habitat. To assure that enough suitable nest sites remain available, the typical nest characteristics and distribution must be known and understood; this is particularly important for an endangered species such as this, which has a restricted natural distribution of 187 hectares on a single, remote island.

METHODS

In all three field seasons from 1986 through 1988, the plant substrate for each nest was recorded, as well as the vegetation type in which the substrate occurred. Based on Newman's (1988) and Lamoureux's

(1963) vegetation classifications, I recognized five vegetation types, which were similar to, but not identical with, either of their classification systems (see Chapters I and IV). The predominant vegetation associations are the Eragrostis variabilis bunchgrass association (Newman's Eragrostis Grassland and Eragrostis Mix, and

83 Lamoureux's Eragrostis Association) and the Ipomoea-Boerhavia-Sicvos viney association, (Newman's Ipomoea Dominant and Sicvos associations,

and Lamoureux's Boerhavia-Ipomoea-Tribulus Association). Eragrostis also occurs as a subdominant in some, but not all, areas of the Ipomoea­

Boerhavia-Sicvos association. Based on regular feeding observations made concurrently in all five vegetation associations in all three years, I found that the vast majority of the nests occurred in these two vegetation types, which I will refer to as "bunchgrass" or "viney", respectively. The few nests that occurred outside these two major vegetation associations, or occurred at the border between two or more vegetation associations, were classified as "mixed". Due to the scarcity of nests outside the bunchgrass and viney associations, I concentrated my 1987 and 1988 efforts in a study area containing only those two habitat types. Except for a few of the early 1986 nests, which were in the northeastern area of the island, all nests in this study were located in the primary study area (Figure 2-1) on the northwestern side of the island, where I confined the majority of my searching.

Nest Characteristics

During the 1987 field season, detailed observations were made on the first 85 nests or nest-like formations that I located in my study area. These are henceforth referred to as "intensely described nests".

Only 68 of these 85 nests were eventually finished and contained eggs; I have excluded the other 17 nests from all the analyses. Data on the 68 active nests included: visual estimate of percent canopy of each plant

84 species in the composite nest substrate, size (maximum height and width in em) of the composite substrate, number of ~ variabilis clumps in the substrate, inner depth and diameter (in em) of the nest cup, total height and diameter (in em) of the nest, and height of the nest bottom above the ground (in em). In cases where a nest was suspended over a seabird nest burrow, the height was measured from the bottom of the nest to the floor of the hole.

The orientation of a nest in relation to the center of the plant substrate was recorded. For example, blades of Eragrostis almost always lean away from the prevailing winds from the northeast, and nests found beneath that overhang were given a southwest heading. Nests that occupied the interior of erect clumps were assigned the orientation category of "None".

The elevation of the nest within the plant substrate was assigned to one of four categories: on the ground, or in the lower, middle, or upper third of the substrate.

The percent cover immediately over the nest cup was estimated on the morning a nest was discovered, or on the following morning for nests found in the afternoon. A hand mirror marked with a 5 em X 5 em grid was placed directly over the nest cup and the number of square centimeters not reflecting incoming light was multiplied by 4 to get an estimate of percent cover over the nest cup. Except for this measurement, the rest of the nest measurements were made either after the eggs hatched but before the chicks were banded (usually when they were 11 to 14 days old), or as soon as possible after a nest failed

(i.e. after eggs disappeared or chicks died).

85 Nest characteristic variables were examined with multiple analysis of variance and were compared for significant differences between the bunchgrass and viney associations using t-tests or Mann-Whitney tests implemented with routines from the PC SAS Version 6.03 software package on a personal computer (SAS 1988).

Nearest-neighbor Distances and Nest Density Estimates

During the 1987 and 1988 field seasons, scaled maps were made of locations of all nests known to have had eggs. Both of these nest maps were constructed using magnetic compass headings between nests and distances measured with a metric tape. The only concrete pillar on

Laysan (a U. S. Geological Survey bench mark) was the origin point for each map. Since Laysan is relatively flat, with a maximum elevation of approximately 10 meters and variation of only a few meters over the entire study area (Ely and Clapp 1973), elevational corrections were not made during map construction.

Maps of simultaneously active nests were constructed for 1987 and

1988. A nest was considered active from the day the first egg was laid until the last chick left the nest, or until the nest failed. If a nest was discovered with chicks, the nest's active period was calculated backward to the egg stage based on the apparent age of the chicks according to their development. For each year, the day with the most simultaneously active nests was selected as the peak day of breeding.

The distances between nearest active nests on that day was used to calculate nearest-neighbor distances (Clark and Evans 1954) within the two main vegetation types, as well as average nearest-neighbor distance

86 for both vegetation types combined. The few nests categorized as being

from a mixed vegetation association were lumped into the same vegetation

category as their nearest neighbor.

Using the scaled maps, a grid of 10 m X 10 m squares (0.01 ha per

square) was superimposed over the study area. Based on the locations of

the concrete pillar, the nests, and the main trail, the total area I

searched intensely at the study site during each year was estimated in

hectares. For each year, the number of simultaneously active nests within each vegetation association in the intensely searched area were used to calculate the peak density within each association. As before,

the few nests categorized as being in a mixed vegetation type were

lumped into the same category as their nearest neighbor. Calculations

of nest density were based on the assumption that all simultaneously

active nests within the intensely searched area were located. This

assumption was later shown to be not completely accurate, based on the

ages of some nestlings that were later located in the searched area.

Using a Bryan Modified Acreage Grid (Mosby 1980) on Newman's (1988)

1984 vegetation composition map, I estimated the total hectares on

Laysan for both the Eragrostis and the Ipomoea-Boerhavia-Sicyos

vegetation associations, since such estimates were not otherwise

available.

Clutch Size and Fledging Success

The number of eggs per nest and fledging success from known size

clutches were compared between the two main vegetation types. Fledging

success was defined as the mean number of chicks fledged per nest. A

87 chick was considered fledged if it reached banding age (days 11 to 14) and was not found dead in or near the nest by fledging age (22 to 26 days old). Nests which occurred in a mixed vegetation type were excluded from these analyses.

RESULTS

Nest Characteristics

The nests were constructed mainly of stems, roots, and blades from the bunchgrass Eragrostis variabi1is, although sometimes they contained stems of the non-native plant Cenchrus echinatus.

Of the 68 intensely described nests from 1987, 65 had Eragrostis, one had g. echinatus, and two had Ipomoea pes-caprae as the nest-site substrate. The three nests with non-Eragrostis substrates had

Eragrostis as a significant component on or mixed in their primary substrate. Twenty-two nests had two or more plant species making up part of the substrate or canopy (Table 2-1). When bunchgrass was the primary plant substrate, but other plants were mixed in or on it, the morning glory vine (~ pes-caprae) was the most common (17 of 19 instances) secondary plant. The introduced weed ~ echinatus, and the native cucurbit vine, Sicyos maximowiczii, also occurred as primary or secondary substrate plants. In many instances the nest site was a composite of several species of plants and/or several individual plants of the same species. Nests were hidden from view within the bunchgrass clumps, and were almost never (less than 5% 6f the time) visible without manually searching through the clumps.

88 Eragrostis clumps are the preferred nest substrate (Morin and

Conant 1990), and almost all nests in all three years occurred in the

Eragrostis or the Ipomoea-Boerhavia-Sicyos vegetation associations where the bunchgrass primarily occurs (Table 2-2). While collecting feeding observations in the other three vegetation associations (Pluchea,

Scaevola, and Sesuvium-Heliotropium-Cyperus associations) outside my primary study area, I also searched for nests with little or no success.

Of the 68 intensely described nests in 1987, 46 (67.6%) were in the bunchgrass association, 18 (26.5%) were in the viney association, and 4

(5.9%) were at the border of the bunchgrass and viney association and were classified as "mixed". These four mixed association nests had their nearest active neighbors in the bunchgrass type.

Eragrostis grass blades most frequently lean in the southwest direction due to prevailing winds, and the nests tended to be beneath the bent blades. For the 37 nests where I recorded nest orientation, 15

(40.5%) had a southwest orientation (Table 2-3). Nest orientations were not random with respect to compass direction (Chi Sq = 53.25, d.f. = 8,

P < .001).

The elevation of a nest within its plant substrate was also not random (Chi Sq = 25.6, d.f. = 3, P < .001). For the 44 nests where I recorded nest elevation, 4 were located in the upper third, 10 in the middle third, and 25 in the lower third of their substrate. Another 5 nests were on the ground within their nest substrate.

The 13 other nest characteristic variables are defined in Table 2-4 and their means and standard deviations are presented in Table 2-5.

Values of these quantitative nest characteristics were compared between

89 the bunchgrass and viney vegetation associations for the 68 intensely described nests from 1987. For this analysis, mixed nests were included with the bunchgrass nests, since they were the nearest active neighbors.

A multiple analysis of variance test (MANOVA) indicated that the vegetation association had a significant effect on the character measures (Wilks' Lambda = 0.372, P = 0.0004). Only four variables,

MAXIMUM HEIGHT of the composite substrate, PERCENT CANOPY1, CANOPY OVER

CUP, and NEST WIDTH were significantly different between nests occurring in the two vegetation types (Table 2-5).

Nearest-neighbor Distances and Nest Densities

In 1987, the peak of simultaneously active nests (n=66 nests in 45 pairs) occurred on 15 June, and in 1988 the peak (n=51 nests in 31 pairs) occurred on 26 May (Figures 2-2 and 2-3, respectively). For both years, nearest-neighbor distances for simultaneously active nests were calculated for the bunchgrass and the viney vegetation types, as well as the two vegetation types combined. Using the Dixon test for outliers

(Sokal and Rohlf 1981), one 1988 active nest was excluded as an outlier from the analyses, based on its very large nearest-neighbor distance.

In both years, the average nearest-neighbor distance between nests in the bunchgrass association was significantly greater than the average nearest-neighbor distance in the viney association (Table 2-6). In

1987, t = 4.346, d.f. = 40.6, P=.0001i in 1988 t = 3.832, d.f. = 29, P

=.0006, and for both years combined t= 6.328, d.f. = 71, P=.0001).

The average nearest-neighbor distance for combined habitats in 1987 tended to be smaller than in 1988, but this difference was not

90 significant (t= -1.1004, d.f.=74, P = .2747). Likewise, there was no

significant difference in the average nearest-neighbor distances for the two years in the bunchgrass association (t = -1.569, d.f = 50, P =

.1229) or in the viney association (t = .2961, d.f. =22, P = .7699). An unbalanced two-way ANOVA with year and vegetation association as treatments revealed that the vegetation association, but not the year or the interaction of year and vegetation association, had a significant

effect on the nearest-neighbor distance (based on 76 observations, F =

28.65, P = .0001).

Figures 2-4 and 2-5 show every nest with eggs or chicks that I

observed in the study area during the 1987 and 1988 field seasons,

respectively. In each year, the greatest number of simultaneously

active nests is fewer than the total nests shown on the map, partly due

to non-overlapping renesting attempts. The estimated total area

intensely searched in 1987 was 7.16 ha, and in 1988 it was 8.16 ha. The

viney area searched was a constant 1.28 ha in both years. Table 2-7

summarizes the densities in the two main vegetation types for both

years. In 1987, the nest density in the bunchgrass association was

about one-half that in the viney association; in 1988, the ratio of

bunchgrass to viney nests was even smaller. Thus, the viney association

could be at least twice as productive per unit area as the bunchgrass

association, (but see Fledging Success below).

I estimated from Newman's (1988) vegetation composition map of

Laysan that the entire island had a total of 112.6 ha in the bunchgrass

association and 50.8 ha in the viney association. Using my density

estimates from both years, I calculated minimum estimated numbers of

91 nests at the peak of the breeding season for each vegetation type and for both types combined for the whole island (Table 2-7). I consider these to be minimums because I know from other observations that I did not detect every active nest.

Clark and Evans (1954) showed how the nearest-neighbor distances, r a, could be used as a measure of spatial relationships. If r e is the mean nearest-neighbor distance expected in an infinitely large random distribution of density rho and R = ra/re, then R = 1 in a random distribution, and R = 0 in a maximally aggregated distribution. In

1987, R was .85 for the bunchgrass association, and .67 for the viney association. In 1988, R was .83 for the bunchgrass and .57 for the viney association.

Fledging Success in Relation to Habitat ~

Mann-Whitney tests indicated that there was not a significant difference between average clutch size (using only known size clutches)

in the bunchgrass versus the viney vegetation association during any year (Table 2-8). Nests categorized as mixed vegetation association

were not used in these analyses.

There was not a significant difference in the average number of

fledglings per nest between the two vegetation associations for 1987 or

1988 (Table 2-9). Nests with clutches of unknown size and nests in the

mixed vegetation category were excluded from these analyses. In 1986

there were significantly more fledglings per nest in the bunchgrass

association (1.6 per nest versus 0.7 per nest; Mann-Whitney test, Z=

2.108, P = .035). However, these data were combined from two breeding

92 peaks which I observed that year (see Chapter I); when the two peaks

were examined separately (as "early" and "late"), there was no

significant difference between the fledging success per nest in the two

vegetation associations (Table 2-9). The only fledgling from an early

1986 nest of known clutch size occurred in the viney association. The

few other fledglings seen immediately after the early peak of breeding

in 1986 all occurred in, or next to, the viney plant association. There were few nests in the bunchgrass association in early 1986.

DISCUSSION

Nest Characteristics

A generalized Laysan Finch nest and nest substrate can be described based on my observations of the 1987 subset of intensely described

nests. The average nest had Eragrostis as a primary substrate and

dominant canopy. One-third of such nests had a second plant species mixed in or growing over the nest substrate. The average nest was in the lower or middle third of the substrate. It usually was oriented toward the south, southwest, or west, consistent with the inclination of the substrate, as shaped by the wind. The substrate was usually a

composite of two bunchgrass clumps, and at least one of the clumps was

large and mature. The average maximum height and maximum width of the

composite substrate were a meter and a meter and a half, respectively.

About 83% of the nest cup was shaded in the morning. The average nest

cup was 3.8 cm deep and 7.1 cm in diameter in inside dimensions; the outside of the nest was 6.9 cm in height and 15.8 cm in diameter. The

nest was an average height of 12.3 cm above the ground. These

93 generalizations are consistent with my observations from all three field seasons.

The inner nest cup and outer nest dimensions in this study (Table

2-5) are similar to measurements taken by Crossin (1966) on seven Laysan

Finch nests (averages: 5.3 cm inner cup depth, 7.4 cm inner cup diameter, 6.9 cm outer nest height, and 13.7 cm outer nest diameter).

Van Riper's (1980) measurements on 26 nests of a closely related

Hawaiian honeycreeper, the Palila (Loxioides bail1eui), also are similar to nests from this study (averages: 3.9 cm inner cup depth, 7.4 cm inner cup diameter, 7.7 cm outer nest height, and 14.7 cm outer nest diameter). Neither Crossin nor van Riper identified the ages of nests they measured. Age differences may account for Crossin's generally larger inner nest cup measurements, since the nest cup changes in size and shape as the nestlings age (pers. obs.). The Palila is a larger bird than the Laysan Finch, and a larger nest cup is expected.

Crossin's observations (1966), as well as those of this study, confirm the Laysan Finch's proclivity for nesting in dense vegetation, presumably for thermal buffering (Weathers and van Riper 1982) and possibly for concealment from avian predation (Morin and Conant 1990).

The high percent of cover over the nest cup (CANOPY OVER CUP) and the height of the nest above the ground (NEST HEIGHT) in proportion to the height of the nest substrate (MAXIMUM HEIGHT; Table 2-5) provide such protection and concealment. Drent (1983) recognized that nest orientation could influence thermal regulation: e.g., by facing into or away from the sun, or by facing into or away from prevailing winds, nests may be warmed or cooled. Prevailing winds on Laysan blow from the

94 east and northeast direction. The nests in this study (Table 2-3) tended to be built underneath the southwestern side of overhanging bunchgrass canopies. This orientation provided these nests with shade from morning and noon sun, as well as protection from direct exposure to prevailing winds. In addition, grass blades that bend (in any direction) provide better protection from heavy rainstorms than do erect blades. Kern and van Riper (1984) found that another Hawaiian honeycreeper, the Common 'Amakihi (Hemignathus virens virens) always nested below the surface of the tree's canopy and suggested that the nests were therefore better shielded from rain. The White-browed

Sparrow-Weaver builds its nest on the side of the tree facing away from prevailing winds (Collias and Collias 1984, Ferguson and Siegfried

1989). Ferguson and Siegfried (1989) believed that this positioning allowed the nests to remain intact for longer periods of time, facilitating year-round breeding and roosting. Laysan Finches have a prolonged breeding season, so it is clearly advantageous to position nests so they receive minimal environmental wear. They also roost in

Eragrostis clumps, and although adults have never been observed roosting in old nests, hatch-year (BY) birds have infrequently been seen resting in them.

In a recent study of Palila on Hawaii Island (Pletschet and Kelly

1990), nests in widely spaced, large trees were more successful than nests in closely spaced, smaller trees. The authors suggested that some extrinsic factor (such as predation) made the larger trees better nest sites. They did not discuss the possibility that thermal or mechanical

95 weather effects (e.g. high winds) could heavily influence nest success

in large versus small trees.

For the Laysan Finch data, there appear to be minor differences in

the nest characteristics between nests in the bunchgrass and the viney

associations. One might expect larger, seemingly better insulated

substrates and nests to provide more thermal buffering and hence afford

higher nest success. Counterintuitively, there was no difference in

nest success between the two vegetation types, despite the taller

maximum bunchgrass clump height, higher percent of cover over the nest

cup, higher percent canopy of the primary nest substrate, and wider

outer nest diameter in the bunchgrass association (Table 2-5).

The larger MAXIMUM HEIGHT and CANOPY OVER CUP in nests from the

bunchgrass association probably reflects an overall larger growth form

of bunchgrass in that association, where vegetation in general tends to

be spaced farther apart than it is in the viney association. The larger

PERCENT CANOPY1 in the bunchgrass association is an inevitable result of

the fact that fewer nest substrates in the viney association were 100%

Eragrostis. The larger NEST WIDTH for nests in the bunchgrass

association may simply occur because more nesting material is available

there, especially since nest densities were lower in that association

relative to the viney association. However, these divergent nest

characteristics did not appear to be correlated with measurable

differences in nest success during this study. If the microclimate of

the bunchgrass association generally has a higher temperature due to its

expanses of nonvegetated sand, then possibly the thicker nests and

96 taller, denser-canopied nest substrates are needed in that association

in order to successfully fledge chicks.

Nearest-neighbor Distances, Nest Density, and Fledging Success

Patterns of animal distribution are particularly difficult to analyze, although mapping and nearest-neighbor distance are often used

in an attempt to quantify density and spacing. Warkentin and James

(1988) felt that nest site selection, and hence spacing, could not be understood without understanding the territoriality of a species.

Similarly, Haila (1988) concluded that erroneous determinations of territory and home range in fragmented habitats could have serious consequences on estimates of density. Ripley (1985) considered that analysis of nest patterns was at or beyond the current limits of knowledge due to edge effects and the patchy nature of habitats.

Notwithstanding such pessimism, I have attempted to describe characteristic nest spacing for Laysan Finches using nearest-neighbor distances and densities, acknowledging that all habitats are patchy and heterogeneous at some level. For example, in my study area (Figures 2-4 and 2-5), nests in the viney association are clustered at the eastern side, probably due to an almost total lack of Eragrostis on the western side. However, the average number of bunchgrass clumps per unit area seems to be the same in both the viney and the bunchgrass associations

(Morin and Conant 1990).

Laysan Finches do not defend a traditional territory, but forage away from the nest. Nests are often in close proximity to one other.

In 1988, two simultaneously active nests were only 2.24 m apart. This

97 is very similar to the typical "1-2 m between clumps" that Newman (1988) reported in her description of Eragrostis clump distribution. The R values (Clark and Evans 1954) of .67 and .57 for the viney association, and .85 and .83 for the bunchgrass association in 1987 and 1988, respectively, suggest that the nests are slightly more aggregated in the viney association. However, in both plant associations nests are more randomly spaced rather than aggregated. The tendency toward aggregation in the viney association may simply reflect the spatial distribution of appropriate nest substrates (Le. bunchgrass clumps). Rendell and

Robertson (1989) have suggested a similar explanation for the spacing of secondary hole-nesting birds. They thought that the pattern of nest placement in natural habitats could be partly a result of the spatial dispersion of natural cavities.

Nearest-neighbor distances and density estimates from this study suggest that the viney association on Laysan usually has at least twice as many nests per unit area as the bunchgrass association (Tables 2-6 and 2-7). Why should the viney area have a higher nest density and appear to be a preferred vegetation type for nesting, even though the average clutch size per nest (Table 2-8) and the average number of fledglings per nest (Table 9) were not significantly different in the two vegetation associations? There are several possible explanations for the apparent preference for the viney association. First, the proximity of good foraging areas may influence nest site selection, allowing pairs to nest more densely and thus expend less time and energy foraging for food. The western side of the viney area at my study site was a popular finch foraging ground, partly because the native cucurbit 98 vine (~maximowiczii) and other food plants were usually abundant. The

Sicyos fruit, as well as the numerous invertebrates on Sicyos and

Ipomoea (morning glory) leaves, seemed to be an important source of food for nestlings (Morin pers. obs.).

Secondly, the viney association may be preferred because some characteristic of the habitat leads to increased fledging success there during breeding peaks very early or late in the season under marginal weather conditions (see Chapter I). I was able to witness an early breeding peak only in 1986. After that peak, the viney area was the only vegetation association where a few fledglings were found. The few nests I found in the bunchgrass association during that early season produced no young. The topography of the island somewhat protects the innermost rings of vegetation around the lake (e.g. viney association) from the typical heavy wind and rain of winter storms; this may explain the differential nesting success (Table 2-9). Over time, selection would favor birds that nested in the more protected area.

A third possible explanation for the observed nest spacing is habitat constraint, such as availability of preferred nest sites

(Rendell and Robertson 1989, Hagan and Walters 1990). However, the density of bunchgrass clumps is thought to be similar in the two associations, and few of the substrate measurements were significantly different. The absolute density of apparently suitable bunchgrass clumps seems high enough to be nonlimiting; on average there are approximately 35 bunchgrass clumps available to each Laysan Finch (Morin and Conant 1990). This suggests that other factors, such as the two mentioned above, may be more likely causes for the nest distribution.

99 However, if the topography of the viney area is more protective, or foraging areas in or near the viney area are better, then nest sites may in fact be limited in the viney area. In that case, at some threshold density, finches may be at a selective advantage if they nest in the less preferred bunchgrass association, especially if higher nest densities in the viney association facilitate predation on finch eggs.

The apparent decline in nest densities between 1987 and 1988 (Table

2-7) is concordant with the reduced finch population estimated during that same time period (see Chapter IV). In 1988 the peak of simultaneously active nests occurred on 26 May, only two days before the breeding census. I have postulated elsewhere that during the peak of the nesting period, females are undercounted during censuses because only females incubate and brood. I have also proposed (see Chapter IV) that census estimates from the peak breeding season be multiplied by the rough correction factor of 7/5, to compensate for incubating and brooding females. The 1988 breeding population estimate was 5,201 finches, and the corrected population for that date (28 May) would therefore be 7,281 finches. This corrected number also represents a pre-breeding population estimate. The difference between the corrected estimate (7,281) and the direct census estimate (5,201) was 2,080 finches, which may be taken to represent the number of breeding females that were incubating or brooding and not counted during the breeding census. If this is true, then these 2,080 uncounted females had 2,080 counted mates, and about 4,160 finches (57%) had active nests in the incubation or brooding stage on 28 May 1988 during the peak nesting period. This suggests that about 3,121 finches (43%) were not breeding.

100 The non-breeders may be second-year birds that are too young to breed or mature birds that bred earlier in the year or will breed later in the year. My island-wide estimate for simultaneously active nests on 26 May

1988 was 1,231 nests, which is much less than the uncounted 2,080 females from the corrected estimate. This suggests that in my study area I had located only about 59% (1231/2080) of the active nests by 28

May, which I believe to be a reasonable possibility.

In 1987, the peak of simultaneously active nests occurred on 15

June (Figure 2-2), nearly three weeks after the breeding census on 28

May. For this reason, I did not use the correction factor with the 1987 breeding population census to estimate a pre-breeding population size, since it would result in an overestimation.

It is apparent that at the peak of breeding in both years, the much smaller area of viney habitat (an estimated 50.8 ha) is as important for overall potential finch recruitment as the larger bunchgrass habitat (an estimated 112.6 ha; Table 2-7). The estimated number of total nests on the day of peak nest activity was almost the same in the two vegetation types, even though the total hectares of viney habitat is only half as large as the bunchgrass association.

Results of this study indicated few differences between nest sites in the viney and bunchgrass associations. It remains unclear why finch nest counts and nearest-neighbor distances from 1987 and 1988 indicated higher densities in the viney association, even though average clutch size and observed fledging success per nest were not significantly different between the two vegetation types.

101 Given the importance of the viney association to overall finch nesting and recruitment, the rapid spread of the non-native Pluchea indica shrub over the past 29 years along the lake shore (Newman 1988) is of concern. This shrub apparently has invaded and is currently invading areas that were previously viney habitat. No Laysan Finch nest has ever been reported in a live Pluchea shrub. The rapid spread of this plant into the important nesting habitat for the endangered Laysan

Finch certainly deserves further scrutiny. Long-term protection of the nesting habitats, especially the viney association, from non-native plants should be a priority action for preserving this endangered bird.

102 LITERATURE CITED

Burger, J. and M. Gochfeld. 1988. Nest-site selection by Roseate Terns

in two tropical colonies on Culebra, Puerto Rico. Condor 90: 843-

851.

Clark, P. J. and F. C. Evans. 1954. Distance to nearest neighbor as a

measure of spatial relationships in populations. Ecology 35: 445­

453.

Collias, N. E. and E. C. Collias. 1984. Nest building and bird

behavior. Princeton University Press, Princeton, New Jersey. 336 pp.

Crossin, R. S. 1966. Notes on the Laysan Finch (Psittirostra cantans).

Leeward Island Survey No. 13. Unpubl. Report, POBSP, Smithsonian

Institution, Washington, D.C. 10 pp.

Drent, R. 1983. Incubation. Pp. 333-420 In Avian biology. Vol.

VII. Academic Press, Inc.

Ely, C. A. and R. B. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll Res. Bull. No. 171.

Smithsonian Institution, Washington, D. C. 361 pp.

Ferguson, J. W. H. and W. R. Siegfried. 1989. Environmental factors

influencing nest-site preference in White-browed Sparrow-weavers

(Plocepasser mahali). Condor 91 (1): 100-107.

Hagan, J. M. and J. R. Walters. 1990. Foraging behavior, reproductive

success, and colonial nesting in ospreys. Auk 107: 506-521.

Haila, Y. 1988. Calculating and miscalculating density: the role of

habitat geometry. Ornis Scand. 19: 88-92.

103 Kern, M. D. and C. van Riper III. 1984. Altitudinal variations in

nests of the Hawaiian honeycreeper Hemignathus virens virens. Condor

86: 443-454.

Kerpez, T. A. and N. S. Smith. 1990. Nest-site selection and nest­

cavity characteristics of Gila Woodpeckers and Northern Flickers.

Condor 92 (1): 193-198.

Lamoureux, C. H. 1963. The flora and vegetation of Laysan Island.

Atoll Res. Bull. No. 97. Nat. Academy of sci., Nat. Res. Council,

Washington, D. C. 14 pp. + figures.

Leonard, M. L. and J. Picman. 1987. Nesting mortality and habitat

selection by Marsh Wrens. Auk 104: 491-495.

McAuliffe, J. R. and P. Hendricks. 1988. Determinants of the vertical

distributions of woodpecker nest cavities in the Sahuaro cactus.

Condor 90: 791-801.

McCallum, D. A. and F. R. Geh1bach. 1988. Nest-site preferences of

Flammulated Owls in western New Mexico. Condor 90: 653-661.

Morin, M. P. and S. Conant. 1990. Nest substrate variation between

native and introduced populations of Laysan Finches. Wilson Bull.

102: 591-604.

Morin, M. P. In prep. The breeding biology of an endangered Hawaiian

honeycreeper, the Laysan Finch (Telespiza cantans).

Morin, M. P. In prep. Laysan Finch population estimates in relation to

the annual breeding cycle and other variables.

Mosby, H. S. 1980. Reconnaissance mapping and map use. Pp. 277-290.

In Wildlife management techniques manual. 4th ed. S. D. Schemnitz

(ed.). Wildl. Soc., Washington, D. C.

104 Newman, A. L. 1988. Mapping and monitoring vegetation change on Laysan

Island. M.A. Thesis, Geography Dept., Univ. of Hawaii, Honolulu.

234 pp.

Pletschet, S. M. and J. F. Kelly. 1990. Breeding biology and nesting

success of Palila. Condor 92: 1012-1021.

Rendell, W. B. and R. J. Robertson. 1989. Nest-site characteristics,

reproductive success, and cavity availability for tree swallows

breeding in natural cavities. Condor 91: 875-885.

Ripley, B. D. 1985. Analysis of nest spacings. Pp. 151-158. In

Statistics in ornithology. B. J. T. Morgan and P. M. North (eds.).

springer-Verlag, New York.

SAS Institute Inc. 1988. SAS/STAT User's Guide, Release 6.03 Edition.

Cary, NC: SAS Institute Inc. 1028 pp.

Sincock, J. L. and E. Kridler. 1977. The extinct and endangered

endemic birds of the Northwestern Hawaiian Islands. unpubl. Report,

U. S. Fish and Wildlife service, Honolulu. 111 pp.

Sokal, R. R. and F. J. Rohlf. 1981. Biometry. W. H. Freeman and

Company, New York. 859 pp. van Riper III, C. 1980. Observations on the breeding of the Palila

Psittirostra bailleui of Hawaii. Ibis 122: 462-475.

Warkentin, I. G. and P. C. James. 1988. Nest-site selection by urban

Merlins. Condor 90(3): 734-738.

Weathers, W. W. and C. van Riper III. 1982. Temperature regulation in

two endangered Hawaiian honeycreepers: the Palila (Psittirostra

bailleui) and the Laysan Finch (Psittirostra cantans). Auk 99: 667­

674.

105 TABLE 2-1. Primary plant substrate (with secondary substrate and canopy plants listed in order of decreasing importance) for intensely described nests from 1987.

SUBSTRATE (SECONDARY PLANTS) FREQUENCY

Eragrostis variabilis (no secondary plants) 44

~ variabilis (unknown whether secondary plants included) 2

~ variabilis (Ipomoea pes-caprae) 15

~ variabilis (~pes-caprae and Sicyos maximowiczii) 2

~ variabilis (~maximowiczii) 1

~ variabilis (Cenchrus echinatus) 1

~ pes-caprae (~ variabilis) 2

~ echinatus (~variabilis) 1

TOTAL 68

106 TABLE 2-2. Number of nests in 1986, 1987, and 1988 in each vegetation classification (and subset of nests with known size clutches).

YEAR BUNCHGRASS VINEY MIXED

1986* 26 (20) 56 (44) 12 (11)

1987 57 (27) 27 (18) 4 (4)

1988 64 (22) 27 (4) 5 (2)

* Includes nests from both the early and late breeding pulses.

107 TABLE 2-3. Frequency of nest orientation.

COMPASS ORIENTATION FREQUENCY

North o

Northeast o

East 2

Southeast 1

South 5

Southwest 15

West 10

Northwest o

No Orientation 4

TOTAL 37 NESTS

108 TABLE 2-4. Nest characteristic variables and definitions.

PERCENT Percent of canopy composed of the nest's primary substrate CANOPYl plant.

PERCENT Percent of the canopy made up of the most abundant plant other CANOPY2 than the primary substrate plant.

TOTAL Total number of bunchgrass clumps contiguous with the primary CLUMPS nest substrate, that can be considered part of the composite substrate.

MAJOR Number of major (mature) bunchgrass clumps in a composite CLUMPS substrate.

MINOR Number of immature bunchgrass clumps in a composite substrate. CLUMPS

MAXIMUM Maximum height of the composite substrate (cm). HEIGHT

MAXIMUM Maximum diameter of the composite substrate (cm). WIDTH

CANOPY Percent of cover over the nest cup during the morning. OVER CUP

CUP Inner depth of the nest cup (cm). DEPTH

CUP Inner diameter of the nest cup (cm). WIDTH

NEST Total height of the nest (ern). DEPTH

NEST Widest outer diameter of the nest (ern). WIDTH

NEST Height of the bottom of the nest above the ground (em). HEIGHT

109 TABLE 2-5. 1987 nest characteristic variables* for both vegetation associations, separately and combined, and results of t-tests or Mann­ Whitney tests between the two associations.

VARIABLE* COMBINED BUNCHGRASS VINEY TEST P PARAMETER

PERCENT MEAN 90.38 % 95.61 % 75.29 % Z = 0.0001 CANOPYl S •D. 19.24 14.99 22.46 -5.726 N 66 49 17

PERCENT MEAN 28.41 % 35.00 % 25.94 % Z = 0.374 CANOPY2 S.D. 23.72 28.81 22.08 0.890 N 22 6 16

TOTAL MEAN 1.89 2.02 1.53 Z = 0.109 CLUMPS S.D. 0.96 1.03 0.62 -1.604 N 66 49 17

MAJOR MEAN 1.35 1.41 1.18 Z = 0.208 CLUMPS S.D. 0.59 0.64 0.39 -1. 259 N 66 49 17

MINOR MEAN 0.55 0.61 0.35 Z = 0.368 CLUMPS S.D. 0.83 0.89 0.61 -0.900 N 66 49 17

MAXIMUM MEAN 90.77 cm 93.69 em 82.35 em t = 0.025 HEIGHT S.D. 18.15 18.43 14.73 2.292 N 66 49 17

MAXIMUM MEAN 149.83 em 153.69 cm 138.71 cm t = 0.201 WIDTH S.D. 41.45 41.66 39.98 1.291 N 66 49 17

CANOPY MEAN 82.75 % 85.73 % 74.35 % Z = 0.033 OVER CUP S.D. 17.05 14.68 20.69 -2.137 N 65 48 17

CUP MEAN 3.83 cm 3.80 em 3.92 em t = 0.684 DEPTH S.D. 0.85 0.87 0.85 -0.410 N 44 32 12

CUP MEAN 7.08 cm 7.13 em 6.94 em t = 0.467 WIDTH S.D. 0.76 0.73 0.85 0.733 N 44 32 12

* Variables are defined in Table 2-4.

110 TABLE 2-5 (Continued)

VARIABLE COMBINED BUNCHGRASS VINEY TEST P PARAMETER

NEST MEAN 6.88 em 6.91 em 6.82 em t = 0.867 DEPTH S.D. 1.60 1.71 1.35 0.169 N 44 32 12

NEST MEAN 15.78 em 16.40 em 14.11 em t = 0.008 WIDTH S.D. 3.39 3.65 1.77 2.786 N 44 32 12

NEST MEAN 12.26 em 11.64 em 14.03 em Z = 0.379 HEIGHT S.D. 9.24 9.12 9.78 0.879 N 46 34 12

111 TABLE 2-6. Nearest-neighbor distances (meters) for simultaneously active nests in two vegetation associations, separately and combined.

YEAR COMBINED BUNCHGRASS VINEY

1987 MEAN 13.11 m 15.29 m 8.29 m

S.D. 6.88 6.86 3.88 N (PAIRS) 45 31 14

1988 MEAN 15.10 m 18.60 m 7.77 m

S.D. 8.87 8.25 4.77 N (PAIRS) 31 21 10

112 TABLE 2-7. Density of simultaneously active nests in the study area for two vegetation types, and estimated total active nests island-wide on date of breeding peak.

DATE COMBINED BUNCHGRASS VINEY

15 JUNE 1987

SIMUL. ACTIVE NESTS 66 45 21

HECTARES IN STUDY AREA 7.16 5.88 1.28

NESTS/HA IN STUDY AREA 7.65 16.41

ESTIMATED HA ISLAND-WIDE 163.4 112.6 50.8

ESTIMATED ISLAND-WIDE 1695 862 833 NUMBER OF NESTS

26 MAY 1988

SIMUL. ACTIVE NESTS 51 34 17

HECTARES IN STUDY AREA 8.16 6.88 1.28

NESTS/HA IN STUDY AREA 4.94 13.28

ESTIMATED HA ISLAND-WIDE 163.4 112.6 50.8

ESTIMATED ISLAND-WIDE 1231 556 675 NUMBER OF NESTS

113 TABLE 2-8. Mean clutch size in both vegetation types for 1986 to 1988 and results of Mann-Whitney tests for differences.

YEAR BUNCHGRASS VINEY Z VALUE P

1986 MEAN 3.4 3.0 1.6278 0.1036

S.D. 0.75 0.70 N 20 44

1987 MEAN 3.0 3.1 0.4217 0.6733

S.D. 0.58 0.82 N 28 18

1988 MEAN 3.5 3.75 0.3817 0.7027

S.D. 0.80 0.96 N 22 4

114 TABLE 2-9. Average number of fledglings per nest (known-size clutches only) in each vegetation association in all three years and Mann-Whitney test results. Nests categorized as "mixed" association were not used.

YEAR BUNCHGRASS VINEY Z VALUE P

EARLY 1986 MEAN 0.00 0.05 -0.3814 0.7029

S.D. 0.0 0.213 N 5 22

LATE 1986 MEAN 2.13 1.41 1.5416 0.1232

S.D. 1.506 1.297 N 15 22

1987 MEAN 1.04 0.44 -1. 6149 0.1063

S.D. 1.170 0.705 N 28 18

1988 MEAN 0.64 1.50 1.5078 0.1316

S.D. 1.093 1.291 N 22 4

115 Ephemeral lake --_..- ... . -

Lake and mudflat

N

~ Vegetated Area

o I

Figure 2-1 Map of Laysan Island showing primary study area. 116 70 65 60 (/) 55 t- ~ 50 z w 45 > t- 40 ~ 35 ~ 30

0:=w 25 ~ 20 ~ 15 10 5 O~~"""""""""'~...... ,rw-""""""""''''''''''''''''''''''''''''IoI.I.Io~..J.IooIoIIoI.I.Io~.w- MAY 17 JUN 6 JUN 26 JUL 16 DATE

FIGURE 2-2 Number of simultaneously active nests in the study area during the 1987 field season.

117 55 50 45 ....(/) (/) w 40 z w 35 >.... o 30 -c I..L. 25 0 0::: 20 w m ~ 15 ::::> z 10 5 0 APR 22 MAY 12 JUN 1 JUN 21 JUL 11 DATE

FIGURE 2-3 Number of simultaneously active nests in the study area during the 1988 field season.

118 -100M -50M OM 50M 100M 150M 200M -I---~-I----....j.----l~----I----""'----l

... A

A ... A .... A A A ...A ...... A ... 6.A A I ... en .. I o ~ - ... L A .. * pillar ... Ril viney ~ area o o ~ ... A en o ~ A

I A tv - o o ...... ~ ...... I ... tv en -1-----+---1-1-- -1-- - - - +------1- 1....---+------1 o s::

I tN o l o ...... ~ ...... 1\ ."/"/'v~":<,,x,./,,./,,/,,/',/'/r,,"xx /,/,/, .,,/" ". "" l­ en >( I w tN en == o ~ NORTH>

I -I>­ o o s:: FIGURE 2-4 1987 nest map from Laysan study area. 119 -100M -50M OM 50M 100M 150M 200M • • .. • • • • • •• I" • • • • • • • •• • I l- • • U1 o s::: L • pillar •• * ,.. • f2j viney 4 • area o • 1-, o s::: L IL • !- '-

I N o l o s::: I.- -- ~ • • • I l.- I N • (}l o s::: • • •• l.- i 10._ • •••• I 0 VJ .~ o .- • • o • • • • .4 • s:::: • I • • ••• • I-" xxxxx xx •.;< x x Xx x x Xx' U) Y, x,x:.'-,,&--:~ x: 7x;yv0 .x '-: XX:~: ~~~/"'~>X~.>-X(,X Y·'X~«y ' . I IVv~~" .j:>.. !><~.< o o s::: FIGURE 2-5 1988 nest map from Layson study area. 120 CHAPTER III

GROWTH AND MORTALITY OF LAYSAN FINCH (TELESPIZA CANTANS) NESTLINGS

IN RELATION TO HATCHING ASYNCHRONY

INTRODUCTION

Hatching asynchrony is a well-documented and common phenomenon in

many species of passerines, as well as raptors, the Ardeidae, and the

Larinae (Howe 1978, Clark and Wilson 1981, Mock 1984, O'Connor 1984,

Mead and Morton 1985). Hatching asynchrony generally occurs when

incubation begins prior to the laying of the final egg in a clutch. The

eggs are then different ages developmentally, and they usually hatch in

the laying order. The range from completely synchronous to completely

asynchronous hatching is a continuum, and the definition of asynchrony may vary among researchers.

Reported here are the results for three seasons of research of

chick growth and survival in the asynchronously hatching Laysan Finch.

This endangered species is the sole passerine on a remote, uninhabited

Pacific refuge island without mammalian or reptilian predators. The

island is desert-like, and is subject to a wide range of unpredictable

weather conditions (see Chapter I). My study objectives were: to

describe normal growth for known-age chicks, to show which measurements

were the best predictors of age (i.e. the most conservative), to examine

whether hatching asynchrony affected chick growth and ultimately

survival, and to test some assumptions and predictions of how well

various hypotheses about hatching asynchrony may explain the selection

pressures maintaining asynchrony in this population. I also studied the

121 presence of females at the nest before and during egg-laying, and during incubation.

Hatching Asynchrony Hypotheses

Many hypotheses have been proposed to explain hatching asynchrony in birds. These hypotheses are not necessarily mutually exclusive (Mock

1984, Mead and Morton 1985, Slagsvold 1990). They include: 1) reduced risk of nest predation (the nest failure hypothesis of Clark and Wilson

1981, 1985), 2) adaptive brood reduction via starvation or siblicide which facilitates partial brood success relative to unpredictable resources (Ricklefs 1976, Hahn 1981, Edwards and collopy 1983, O'Connor

1984, Richter 1984), 3) physiological constraints (Mead and Morton

1985), 4) "insurance" eggs whereby excess eggs are a backup if the earlier egg(s) fail (Kepler 1969, Mock and Parker 1986), 5) reduction of peak food demand by nestlings by spreading individual peak needs over a longer time period (Hussel 1972), 6) reduction of sibling rivalry (Hahn

1981, Mock and Ploger 1987), 7) the risk of parental mortality due to predation both at the nest and between nesting attempts (Magrath 1988), and 8) equal parental investment (in dimorphic species) in both sexes of progeny (Slagsvold 1990).

The brood reduction hypothesis (Lack 1954) holds that at least partial brood success can be achieved in the event of an inadequate food supply if the brood size can be adjusted downward by sacrificing the youngest, smallest chicks, often by starvation or siblicide. This hypothesis assumes that there are unpredictable environmental conditions that sometimes lead to food shortages. The hypothesis predicts that 1)

122 if any chicks die due to poor food conditions, later-hatched chicks

should die preferentially (Amundsen and Stokland 1988) in order to minimize food wastage, 2) during poor food conditions, synchronously

hatched broods should be less successful than asychronously hatched broods, 3) the degree of hatching asynchrony may vary between broods or between years in conjunction with food resources, and 4) later-hatched chicks should have lowered growth rates under poor food conditions (Hahn

1981). Other aspects of breeding biology may be associated with

adaptive brood reduction. These could include differential egg provisioning by the female, whereby successively smaller eggs, or a

final small egg are laid (Parsons 1975; Slagsvold et ale 1984). A high

growth rate, which maximizes the "head start" of the earlier chicks, may

also occur (O'Connor 1978a).

The nest failure (predation avoidance) hypothesis assumes that

hatching asynchrony evolved in response to predation at the nest, and

that the risk of predation is some function of the time spent in the

nest. Asynchrony should be adaptive when the daily predation rate on

eggs is equal to or greater than that on chicks (Slagsvold 1988).

Incubation is favored to start on the first egg (Clark and Wilson 1981)

or the penultimate egg (Bancroft 1985) in order to get at least some of

the eggs hatched and out of the nest as quickly as possible. Positive

egg provisioning by the female may occur, with successively larger eggs

laid. This should minimize the inevitable size hierarchy by promoting

equality among the chicks at hatching; egg size has been shown to affect

subsequent chick size (Schifferli 1973). Late in the nestling period,

123 earlier-hatched chicks may have a better ability to escape nest predators.

The physiology hypothesis as described by Mead and Morton (1985) proposed that hatching asynchrony is an incidental effect of hormonal changes during ovulation, causing incubation to begin prior to the completion of the clutch. This hypothesis predicts that incubation should begin with the penultimate egg, regardless of clutch size, simply due to increasing prolactin, which halts ovulation and induces incubation. Egg provisioning, in the form of successively increasing egg size (which occurs in many passerines) may be viewed as a mechanism to reduce the asynchrony. While Mead and Morton do not deny that other advantages may accrue from asynchronous hatching, they view these advantages as epiphenomena.

The insurance hypothesis (Kepler 1969, Mock and Parker 1986), which is a special case of the brood reduction strategy, suggests that extra eggs are laid as insurance for inviable eggs or mortality of ear1y­ hatched chicks. This hypothesis assumes that chick growth should be flexible, so that even a slowly-growing, poorly fed chick can survive a few days as the backup for an older sibling, and that a high rate of infertile or addled eggs, or a high rate of partial (but not total) clutch loss occur. The hypothesis predicts that pairs that lay extra

"insurance" eggs have higher reproductive success than pairs without such eggs.

Hussel (1972) suggested that asynchrony could reduce a clutch's peak food demand by spreading the demand over a longer time period. Few studies have tested this hypothesis, although Fujioka (1985)

124 demonstrated that asynchronous broods of Cattle Egrets ate less food mass, required a lower number of parental visits to the nest with food, and had chicks that begged less often for food than in synchronous broods of the same stage. However, this hypothesis and the sibling rivalry reduction hypothesis (see below) are clearly interrelated.

The reduction in sibling rivalry hypothesis states that hatching asynchrony provides energy benefits to both parents and offspring; clear size hierarchies occur from birth, and broods exhibit lower fighting rates (Mock and Ploger 1987) and higher fledging success (Hahn 1981).

Mock and Ploger (1987) found that synchronous broods were fed more by the parents but did not fledge more offspring, suggesting that considerable energy can be "wasted" in sibling rivalry.

Magrath (1988) proposed that predation on parents could provide selection pressure toward hatching synchrony. He suggested that the probability of the parents being preyed upon before breeding again must be considered in conjunction with the nest failure model.

Recently Slagsvold (1990) proposed that, because of Fisher's sex ratio theory, strongly size dimorphic species should have relatively more hatching asynchrony, since asynchrony would facilitate equal parental investment in progeny where one sex is more "expensive" to raise than the other. This hypothesis assumes that chicks of the more expensive sex should die before chicks of the less expensive sex.

However, this hypothesis does not conflict with several of the other asynchrony hypotheses, and does not provide an explanation for hatching asynchrony in monomorphic species.

125 STUDY AREA AND METHODS

The data were collected during summer field seasons in 1986, 1987, and 1988 on Laysan Island in the Hawaiian Islands National Wildlife

Refuge, Hawaii. Laysan is a low, coral island with open vegetation and very few trees; the vegetation and geography have been described elsewhere (Lamoureux 1963, Ely and Clapp 1973, Newman 1988). Laysan is primarily a seabird rookery, except for a single passerine species, the

endangered Laysan Finch. Prior to 1923, two other now-extinct, endemic

passerines, the Laysan Millerbird (Acrocephalus familiaris familiaris)

and Laysan Honeycreeper (Himatione sanguinea freethii), and an endemic

flightless rail (Porzana palmeri), also inhabited the island.

Most nests of the Laysan Finch were located prior to egg laying.

Nests, and the presence of adults at the nest, were checked daily until

the clutch was completed. Eggs were uniquely marked in the order of

laying, and nestlings were also uniquely color marked according to their

hatching sequence. Nests were checked periodically during incubation,

and every morning, beginning the day before the expected hatching of the

first egg and ending after the last egg hatched or disappeared.

Presence or absence of the incubating female were noted during these

nest checks. A nestling was assigned an age of one day on the day it

was first seen, since it could have hatched any time within the previous

24 hours. Hatching order was recorded for each nestling whenever

possible. Nestlings from clutches where only one egg hatched were

classified as a separate treatment (called "single"). Nestlings were

usually measured every other day until about 11 to 14 days of age, at

which time they received a unique combination of colored plastic leg

126 bands and a numbered u.s. Fish and Wildlife Service aluminum band. To prevent premature fledging, a nestling was usually not measured again after it was banded, except for a few which were fortuitously recaptured as fledglings.

Unflattened right wing chord measurements were made with a metric ruler; right tarsus length, beak depth and lower beak width (at the nares), beak length (straight line length from the nares to the tip), and sternum length were all taken with metric calipers. Except for beak width, these measurements follow conventions outlined in Baldwin et al.

(1931). Occasionally, on tiny one- or two-day-old nestlings, measurements were taken with dividers. Weights were taken to 0.1 g with a 10- or 50-gram Pesola spring scale, depending on the size of the chick. At banding, a chick's furcular fat was recorded on a scale of 0 to 4, where 0 represented no fat, 1 was a trace of fat, 2 was fat on the clavicles and in the concave interclavicular region, 3 was fat filling the furculum more or less level with the pectoralis muscle, and 4 was a convex pad of fat protruding beyond the pectoralis muscle.

A chick was considered to be a fledgling after it was no longer mainly associated with the nest itself, even though it may still have frequented the nest site. Fledging occurred at approximately 22 to 26 days of age (see Chapter I).

Other hatch-year (HY) birds, hatched and fledged within each field season were captured, measured, and banded. Almost all measurements of

HY birds were made during 1988, when I remained on Laysan well past the peak fledging period. Based on field observations of breeding chronology, I know that these HY birds ranged in age from one to at

127 least four months old. In addition to the measurements mentioned previously, the width of the upper mandible (at the nares) and a second beak length (straightline length of the total culmen) were also taken on

HY birds, but are not reported here. In very young chicks, the upper and lower mandible are somewhat soft in texture and are the same width, whereas fledglings have hardened beaks with a lower mandible wider than the upper.

Some of the chick weight data were fitted to the generalized growth

curve as defined by Richards (1959) and modified by Bradley et al.

(1984):

W = A(l + (M - 1) e-KCT - I» 1/(1 - M).

Curve parameters were estimated using an APL program obtained from David w. Bradley (Bradleyet a1. 1984). This program estimated the growth

asymptote (A), a growth constant (K), % of asymptote at inflection (P),

age at inflection (I), time from 10% to 90% of asymptote (G), a curve

shape constant (M), and the weighted mean growth rate (R = KIM). W is

the live body weight at age T.

Data were analyzed using a personal computer statistical software

package (SAS 1988; Version 6.03).

RESULTS

Incubation Pattern

Only female Laysan Finches were observed incubating eggs, and males

do not have brood patches. A female was often found sitting in an

incubating position in an empty nest the day before I found her first

egg. During incubation, the male partner fed the female, generally

128 within 3 m of the nest site (see Chapter I). Females were usually

present on the nest during nest checks, although I frequently flushed them off their nests (Morin unpubl. data).

Hatching Spread

The modal clutch size for the Laysan Finch was three eggs; the

range was from one to five eggs. One-, two-, and five-egg clutches were

rare; four-egg clutches were fairly common (see Chapter I).

Eggs usually hatched asynchronously, with a spread of one to five

days, partly depending upon the size of the initial clutch (Table 3-1).

In a few cases, for each clutch size category, only one egg in the entire clutch survived to hatch. These single chicks do not have a hatching order, and are indicated in Table 3-1 with parentheses. For 10 of the 11 three-egg clutches that hatched during one day, only two eggs hatched in each. Hatching sequence followed the order of laying in all but one nest (see Chapter I).

When the mean weight difference between the last-hatched chick (at

1 or 2 days of age) and the largest (usually first-hatched) chick are

compared for the various clutch sizes (Table 3-2), the size difference

is striking. The mean weight for a typical one-day old chick was 2.89 g

(S.D.=0.52, N=47) and for a two-day old chick was 3.69 g (S.D.=0.85,

N=93; see Appendices A and B). If more than two chicks hatched in a

nest, then the largest chick could be twice as big or bigger than the

last chick to hatch. Later-hatched chicks would seem to be at a considerable competitive disadvantage because of the extreme size differential.

129 ~ Weights in Relation to Laying Order

Overall, bigger clutches tended to have heavier eggs, implying that there is not always a trade-off between clutch size and egg size, but that other factors such as parental quality or condition influence egg weights (see Chapter I). However, there was considerable between-year variation in egg weights in the different clutch sizes. In "good" years of abundant food, big clutches tended to have big eggs, but this was not true in the "bad" year of poorer food availability. For example, four­ egg clutches during a drought year (1987) had significantly lighter eggs than three-egg clutches during that same year; the four-egg weights were more similar to weights from two-egg clutches during that year. This result is consistent with Murphy (1986), who concluded that egg size in

Eastern Kingbirds was a function of female condition more than female

size, and that clutch size was independent of female condition.

If Laysan Finch egg weights are examined in relation to their

laying order, there is a tendency for middle eggs to be heavier than

first or last eggs in three-, four-, and five-egg clutches (Figure 3-1).

Eggs from four-egg clutches in 1987 displayed an exaggeration of this

pattern.

For the modal clutch size (three eggs), a Friedman test was done on

the fresh egg weights within each clutch; laying order was a treatment effect and the nest itself was a factor effect. Both laying order (Chi

Square= 9.9, P<.Ol) and nest effect (Chi square=135.28, P<.OOl) were

significant influences on egg weights (N=67 nests). The middle egg was

usually the heaviest egg of the three.

130 Overall Growth

Growth data were collected from 245 chicks; of these chicks, 54 ultimately did not fledge. Some of the non-fledging chicks were only measured once prior to their death or disappearance. Of the remaining

191 chicks that fledged, I was unsure of the exact age of six chicks; these were excluded from the analyses. A seventh chick was excluded as an outlying data point because of its unusual and inexplicably retarded growth pattern. Although this chick fledged, other chicks that had displayed such signs of retarded growth (Le. "runts") had died within a few days.

The means and standard deviations for the seven growth measurements at various ages in the 184 chicks that fledged are shown in Figures 3-2 to 3-8. The data for the fledged chicks are summarized in Appendix A, and the data for chicks that ultimately did not fledge are summarized in

Appendix B. Appendix A also contains the average measurements for the

240 fledged, HY birds of unknown age. Because chicks take about 24 days to fledge and spend at least another week hiding (and were therefore very unlikely to be captured), I believe that the HY measurements represent birds> 30 days old. However, I know from my field observations that at least some of these HY birds were as old as four months of age. Although the HY beak measurements are slightly less than beak measurements for after-hatch-year (AHY) Laysan Finches, the tarsus, sternum, wing, and weight measurements are indistinguishable from AHY measurements (Conant and Morin unpubl. data). I regard the latter measurements as asymptotic values for HY Laysan Finches (Figures 3-2, 3­

6,3-7, and 3-8). Considerable growth in beak depth, beak length, wing

131 length, and sternum length occurs between banding age and one week (or more) post-fledging (Figures 3-3,3-5, 3-6, and 3-7). For chicks that fledged, the majority of tarsus growth and weight gain is achieved by 15 days of age (Figures 3-2 and 3-7).

I usually measured each individual chick every other day; therefore, adjacent days represent different samples of chicks. I combined measurement data for days 9 and 10, and performed a t-test on the mean measurements for fledged versus non-fledged chicks. The means for fledged chicks 9/10 days old were significantly larger than those for non-fledged chicks for tarsus length, beak depth, beak width, beak length, wing length, and weight. However, the means for sternum length of fledged and nonfledged chicks were not significantly different.

Results are summarized in Table 3-3. Days 9 and 10 were chosen because these days occurred past the inflection point in the growth curves

(Tables 3-4 and 3-5), but before my sample sizes decreased. I rarely measured chicks after they were banded at age 11 to 14 days.

Growth in Relation to Hatching Order

Only chicks of known age and hatching order from known-size clutches of three or four eggs were used in the construction of growth curves and for analysis of variance (ANOVA). An unbalanced three-way

ANOVA with hatch order and clutch size as the treatments, and with

individual broods "nested" as blocks in the clutch treatment to isolate the effect of individual nest environments, revealed that hatching order

(P=.OOl), and the interaction of clutch size with individual brood effects (P=.027), but not clutch size (P=.484) nor the interaction of

132 clutch size with hatching order (P=.3G9), had a significant effect on the weight of a chick at day 9/10 (based on 93 observations, F = 2.39, P

=.002).

using hatch order as the treatment, one-way ANOVAs were made on all seven measurements (at age 9 and 10 days combined) for both three- and four-egg clutches. Three-egg clutches which hatched two and three eggs were analyzed separately (Tables 3-GA and 3-G8), and four-egg clutches which hatched two, three, and four eggs were analyzed separately (Tables

3-7A, 3-78, and 3-7C). Day 9/10 measurements from nests where a single egg hatched are included in Tables 3-GA, 3-68, 3-7A, and 3-78 for comparison only, but were not included in the ANOVA.

For three- and four-egg clutches where only two eggs hatched, there were no significant differences between first- and second-hatched chicks

(Tables 3-GA and 3-7A). For three-egg clutches where all three eggs hatched, only weights were significantly different (Table 3-G8).

Multiple t-tests indicated that the weight of the mean third-hatched chick was significantly less (Alpha=O.OS) than first- or second-hatched chicks, which were statistically indistinguishable.

In four-egg clutches where three eggs hatched, chick beak lengths were significantly different at day 9/10 (Table 3-78). Multiple t-tests

(Alpha=O.05) indicated that beak lengths from first- and second-hatched chicks were statistically indistinguishable, and beak lengths from second- and third-hatched chicks were indistinguishable. The average weight for a third-hatched chick from a four-egg clutch was lower than first- or second-hatched chicks (Table 3-78) and was close to significance (P=O.06). The failure of the ANOVA to distinguish among

133 the weights is probably due to the larger S.D. of the second-hatched chicks. Multiple t-tests grouped the first- and second-hatched chick weights, and the second- and third-hatched chick weights.

There were few four-egg clutches where four eggs hatched, and fewer still for which I had growth measurements. In this subset, beak lengths and chick weights were significantly different based on one-way ANOVA

(Table 3-7C; both P=O.04). Multiple t-tests (Alpha=O.05) grouped the beak lengths for first- and second-hatched chicks, and for first-, third-, and fourth-hatched chicks. Multiple t-tests (Alpha=O.05) grouped the weights of first- and second-hatched chicks, the weights of third- and fourth-hatched chicks, and the weights of second- and third­ hatched chicks. The sample size is small, so these results should be viewed as tentative. They are consistent with the results from four-egg clutches where three eggs hatched (Table 3-7B).

Data were fitted to the Richards' growth model (Richards 1959,

Johnson et al. 1975, Zach et al. 1984) for weights of successfully fledging chicks for each hatching order for both three- and four-egg clutches (Tables 3-4 and 3-5); data are measurements gathered during

1986-1988. The Richards' growth curves for chicks from three-egg clutches where two or three eggs hatched are plotted in Figure 3-9.

Bradley et al. (1984) suggested that G, the time from 10% to 90% of asymptotic size, was probably the best overall growth measurement for comparison purposes. Table 3-4 and Figure 3-9 show the increased G for third-hatched chicks from three-egg clutches. The shape parameter, M, is almost equal to 2 for the first- and second-hatched chicks from three-egg clutches (Table 3-4), indicating that this curve is

134 approximately logistic (Johnson et ale 1975). There is also a large G associated with the fourth-hatched chick in clutches of four; however, the sample size is too small to permit any meaningful interpretation.

Chick Fat in Relation to ~ Weight, Hatching Order, and Mortality

Fresh egg weights were divided into three categories: light, average, and heavy. The mean fresh egg weight was 3.2 g (see Chapter

I); eggs of 3.1 to 3.3 g were categorized as average, 2.4 to 3.0 g as light, and 3.4 to 3.8 g as heavy. There was no association of chick fat at banding age with the chick's fresh egg weight category (N = 101, d.f.

= 4, Chi Sq =3.285, P = .511). For this analysis, only three fat categories (0, 1, and 2) were used. The four chicks with a banding age fat status of 3 were combined into the fat category 2, and there were no chicks with a fat status of 4.

Fat at banding age (days 10 to 15) was significantly associated with a chick's hatching order (N = 127, d.f. = 6, Chi Sq = 18.549,

P=.005). Relatively more chicks in higher fat categories had hatched earlier in the hatching sequence (see Table 3-8). Chicks from nests where only one egg hatched ("single" chicks) were not included in the above analysis. However, when single chicks were combined with first hatched chicks (from nests where more than one egg hatched), the association of fat with hatching order remained significant.

Fat at banding age was significantly associated with subsequent fledging success (N = 286, d.f. = 2, Chi Sq = 10.217, P = .006). Of 286 chicks, 175 of 179 (98%) with a fat status of 2 at banding age fledged,

135 73 of 80 (91%) with a fat status of 1 at banding age fledged, and 23 of

27 (85%) with a fat status of 0 at banding age fledged.

Fledging Success Relative to Hatching Order, Clutch Size, and Year

For a variety of reasons, it is relatively uncommon for all eggs in a clutch to hatch (see Chapter I); this is why my sample sizes became progressively smaller from the first to the last hatched chicks. Only

19% (30 of 158) of the two- to four-egg clutches hatched every egg in the clutch. In contrast, 37% (58 of 158) of two- to four-egg clutches hatched at least one egg (Table 3-9).

For eggs that hatched, fledging success was not independent of the chick'S fresh egg weight (N = 135, d.f. = 2, Chi Sq = 6.416, P. = .04.).

It appears that average-size eggs that hatched had the highest probability of fledging. Twenty-eight of 44 "light" eggs fledged (64%),

29 of 46 "heavy" eggs fledged (60%), and 38 of 45 "average" eggs fledged

(80%) • I have been unable to find a correlation of female weight with her mean egg weight (see Chapter I).

For all chicks where the hatching order as well as the fate

(fledged or not fledged) was known, dependence of hatching order and fate was statistically borderline (N = 169, d. f. = 3, Chi Sq = 7.426, p

= .06). For this analysis, single chicks were not included, and the only fifth-hatched chick was combined into the fourth-hatched category.

Forty-nine of 64 first-hatched chicks fledged (77%), 38 of 54 second­ hatched chicks fledged (70%), 29 of 40 third-hatched chicks fledged

(73%), and 4 of 11 fourth- and fifth- hatched chicks fledged (36%).

136 Single chicks fared better than fourth- and fifth-hatched chick: 12 of

21 (57%) fledged.

Overall, there was superior fledging success for hatchlings from four-egg clutches (Table 3-10). For three-egg clutches where only two eggs hatched, both had an equal probability of fledging. This was also approximately true for four-egg clutches where only two eggs hatched; six of nine first-hatched fledged and five of nine second-hatched fledged. However, for three-, four- and five-egg clutches where three or more eggs hatched, the probability of fledging successfully declined with the hatching order, and appeared to be lowest with the last-hatched chick. For all known-size clutches where three or more eggs hatched, the fate (fledged or not fledged) of the first-hatched and last-hatched chick in a clutch was not independent (N=36 nests, Chi Sq = 5.791, P =

0.016).

The small number of third- and fourth-hatched chicks in four-egg clutches is at least partly due to the large number of addled or infertile eggs. Inexplicably, four-egg clutches had twice as many nests with at least one addled or infertile egg as did three-egg clutches (16 of 40 or 40% versus 18 of 101 or 18%, respectively). For two-egg clutches, 2 of 18 or 11% of the nests had at least one addled or infertile egg. No addled or infertile eggs were discovered in five-egg clutches.

Clutch size was a major determinant of fledging success per nest.

Only one one-egg clutch was seen in all three years of field work, and the chick did not fledge. Of the 18 two-egg clutches, two produced one fledgling each. Although four-egg clutches very seldom produced four

137 fledglings (Table 3-11), relatively more four-egg clutches produced three fledglings when compared to three-egg clutches (Tables 3-11, 3-12,

3-13, and 3-14). An average of 1.35 fledglings per nest (N=40) were produced in four-egg clutches, whereas only 0.85 fledglings per nest

(N=100) were produced in three-egg clutches. A Mann-Whitney test showed that these numbers were significantly different (Z = 2.08, P = 0.038).

Five-egg clutches had variable results, but of the six five-egg clutches seen during the three years, none produced five fledglings.

There also appears to be an effect of year (e.g. due to variable environmental conditions) on fledging success; the modal number of fledglings per nest (in nests where at least one chick fledged) shifted from three fledglings in 1986 to two fledglings in 1987 (Tables 3-12, 3­

13, and 3-14). The 1988 mode was one fledgling per nest, but the small sample size may not represent the true population mode. (In 1988, six nests fledged one chick, four nests fledged two chicks, and five nests fledged three chicks; Table 3-14.)

DISCUSSION

This section will first discuss egg provisioning and chick growth for Laysan Finches in general, followed by an examination of growth and survival patterns in relation to a chick's position in the hatching order. These patterns of egg provisioning, growth, and survival will later be discussed in the context of some of the asynchronous hatching hypotheses.

138 Overall Growth

Chick growth is regulated by several factors, including the ability of the parents to feed the young, the availability and quality of food, the number and competitive ability of sibling chicks, thermoregulatory requirements, and physiological constraints (o'Connor 1984). In general, tropical passerines grow more slowly than do temperate passerines (Ricklefs 1976). Laysan Finches have a lengthy nestling period which most nearly coincides with the protein-limited frugivorous neotropical passerines (Morton 1973), in spite of the fact that Laysan

Finches are omnivorous.

As Appendices A and 8 and Table 3-3 show, chicks that eventually fledged and chicks that failed to fledge (when all three years were combined) followed different courses of growth. Chicks that failed to fledge lagged behind by an average day's increment of growth for every measure except sternum length.

Chick weight proved to be a poor measure for predicting chick age, due to its variability. At nine days of age, chicks that failed to fledge weighed an average of 14.1 g (S.D. = 4.32, N = 8), but nine-day­ old chicks that fledged weighed an average of 20.2 g (S.D. = 2.61, N =

101). At this age, the non-fledging chicks were more similar in weight to seven-day-old chicks that fledged. Although weight, wing length, and tarsus length are often used to predict age (O'Connor 1978b, Ricklefs

1983), my data show that for Laysan Finches, sternum length is probably a good measurement for predicting chick age, since it is conservative and changes insignificantly with nutritional status and hatching order

(Tables 3-3, 3-6A, 3-68, 3-7A, 3-78, and 3-7C; Appendix A and B).

139 Weight and wing length measurements were more variable than sternum

length (their standard deviations were relatively larger, Table 3-3).

Tarsus lengths were relatively less variable than sternum lengths for

chicks that eventually fledged, but relatively more variable for chicks

that did not fledge. Although it is an internal body part, the sternum

of a nestling is easily seen and measured through the thin, un feathered thoracic skin. In contrast, the measurement of adult sternum length is much more difficult, especially for females with brood patches.

~ Provisioning

Intraclutch variation in egg size has been assumed to have adaptive

value (Clark and Wilson 1981, Edwards and Collopy 1983). Regular

patterns in egg size variation with position in the laying order have

been documented. Small final eggs have been thought to facilitate brood

reduction, whereas relatively large final eggs may reduce the effects of

hatching asynchrony by preferentially provisioning the last-hatching egg

(Slagsvold et al. 1984).

Except for two-egg clutches, intraclutch egg weight variation in

Laysan Finches may be consistent with the brood reduction hypothesis, in

that the final egg is relatively small (Figure 3-1). However, the first

egg also tends to be relatively small, which is not predicted by that

hypothesis. Physiological conditions that initiate and terminate egg

laying, as well as the female's energy and nutrient reserves, may better

explain the smaller size of first and last eggs in Laysan Finch clutches with more than two eggs. For some species, a decline in egg mass with

laying order possibly reflects the nutrition of the female; however, the

140 general influence of food supply on egg size and composition is not completely clear (Blem 1990). For example, Murphy (1986) suggested that in Eastern Kingbirds, endogenous lipid or calcium availability may limit egg size.

Growth and Survival Relative to Fat status, Clutch Size, and Hatching

Order

Although growth can be summarized for the "generic" chick, all eggs and their subsequent chicks are not identical due to intrinsic and extrinsic variables. Complex interacting factors influence the growth and survival of any individual egg or chick. It is probable that during different breeding seasons, or different periods within the same breeding season, different factors predominate. I have found previously that during a drought year (1987), average Laysan Finch egg weights were lower and female parents tended to have less furcular fat (see Chapter

I). Schifferli (1973) showed that egg weight had a significant effect on chick weight and growth, although smaller chicks eventually caught up if they survived. In general, eggs are smaller when a female parent has difficulty acquiring enough food (O'Connor 1984).

Although I did not find a significant association of chick fat at banding age with initial fresh egg weight, I did find that fledging success was not independent of egg weight. It appears that selection favored eggs of average weight, since they produced a higher percent of fledglings. However, chick fat at banding age was associated with a chick's hatching order; more of the earlier hatched chicks fell into higher fat categories than later hatched chicks. Fat at banding age was

141 also associated with fledging success: chicks with higher fat status

fledged at higher rates. Since I could not follow chick mortality past

fledging, I have no way of knowing whether chicks with lower fat status

(which comprised a higher percent of the later hatched chicks) had

higher mortality after fledging, although such fledglings are probably

at a disadvantage relative to their fatter peers.

Clutch size was a major factor affecting nests where at least one

chick fledged (Table 3-11). A higher percent of four-egg clutches

produced three fledglings, for all years combined and for each year

taken separately, when compared to two- or three-egg clutches. Two-egg

clutches had very poor fledging success, and five-egg clutches were

inconsistent, apparently varying with unknown year-specific factors such

as food availability. In the drought year (1987), no five-egg clutches

were found, although the search effort was about the same as in the

other two years. There seemed to be a "boom or buat;" production of

chicks in five-egg nests, depending upon the year (Tables 3-12, 3-13,

and 3-14). This supports Boyce and Perrin's (1987) hypothesis that "bad

years" affect individuals laying larger clutches more than they affect

those laying smaller clutches, making it more advantageous in the long

run to lay clutches smaller than the most productive size possible.

However, my sample size (N=6) is too small to draw any statistically

significant conclusions. It is unfortunate that five-egg clutches were

so uncommon, because they might demonstrate most clearly the selective

factor(s) that place an upper limit on clutch size.

In 1986 (Table 3-12), the modal number of fledglings was three in

any nests where at least one chick fledged. However, in the drought

142

------_. year of 1987 (Table 3-13), the mode dropped to two fledglings. In general, for all years (Table 3-11), the four-egg clutches tended to fledge a relatively higher percent in each fledging category (except for nests with two fledglings) relative to two- and three-egg clutches.

This result is in spite of the fact that four-egg clutches were handicapped by a rate of addled/infertile eggs that was twice that of the average three-egg clutch. Arguably, this suggests that parental quality (e.g. age, experience, body condition) is an underlying component affecting both clutch size and fledging success. I have previously shown that female parents with higher fat status had significantly larger clutches and tended to have more fledglings (see

Chapter I).

Chick weight at day 9/10 was affected by hatching order for both three- and four-egg clutches in nests where more than two eggs hatched

(Tables 3-6B, 3-7B, and 3-7C). The first- and second-hatched chicks in nests where three eggs hatched (Tables 3-6B and 3-7B) were very similar, but the third chick fell behind in every measure (although only weight differences were statistically significant, or close to significance).

Hatching order affected the weight of a chick at day 9/10 and also influenced whether it fledged or not. Table 3-10 illustrates how first­ and second-hatched nestlings in nests where only two eggs hatched had similar fledging success, but nests with three or more hatchlings showed some degree of brood reduction. These results support Smith's (1988) conclusion that the brood reduction strategy may be associated only with certain clutch sizes (i.e. the larger or largest clutch sizes). I further suggest that single-year studies that attempt to evaluate the

143 selective advantage of brood reduction may present an incomplete picture. If such a study occurs in a year of "good" food availability, synchronous and asynchronous broods may do equally well, and the result would not conflict with the brood reduction hypothesis; Amundsen and

Stokland (1988) acknowledged this possible limitation to their study.

Although intuitively one might expect that "single" chicks (Tables

3-4 and 3-5) should have high growth rates because of their opportunity to monopolize parental feeding efforts, my results indicate that this is not necessarily true. In three-egg clutches, single chicks were similar to second-hatched chicks for G (the number of days from 10% to 90% of asymptotic weight) as well as R, but this was not true for single chicks from four-egg clutches. Although single chicks in four-egg clutches had a high growth rate, their P (percent of weight asymptote at inflection) was most like a second-hatched chick, and their G was most like a third­ hatched chick. In addition, single chicks fledged at a lower rate than that of third-hatched chicks (58% of 24, versus 73% of 41). This suggests that other factors may strongly influence growth rate, such as parental inexperience, loss of one of the parents, reduced stimulus to parents for feeding, or the lack of sibling brooding.

Overall, from the perspective of the chick, it was best to be in a four-egg clutch. Thirty-four percent of all the eggs in four-egg clutches produced fledglings, whereas only 28% of the eggs from three­ egg clutches did so. The number of fledglings produced per nest for four-egg clutches was on the average better than for three-egg clutches:

55 fledglings per 40 nests vs. 84 fledglings per 100 nests (Table 3-11).

Also, fewer four-egg nests had total nest failure (i.e. no eggs produced

144 fledglings). Because fatter females lay bigger clutches (see Chapter

I), all these data suggest that parental condition is positively correlated with clutch size and nest success under some conditions. It is likely that parental condition is also positively correlated with parental experience or age.

Overall, the best hatch order position (relative to growth and survival) was the first-hatched, the second best was second-hatched, and so on, although the best hatch order position was also influenced heavily by clutch size. The third-hatched chick in four-egg clutches fledged at a higher rate than the first-hatched chick in three-egg clutches; this reinforces that notion that parents of bigger clutches are somehow of better quality (Table 3-10). There was no advantage associated with being a single chick from a three-egg clutch, since such chicks showed a growth rate no better than first or second chicks that grew with siblings (Table 3-4). In addition, single chicks from three­ egg clutches fledged at a rate similar to second-hatched chicks from three-egg clutches (6 out of 9, or 67%). Single chicks from four-egg clutches had a G value (days from 10% to 90% of asymptotic weight) similar to the first three hatchlings in four-egg clutches (Table 3-5).

However, they fledged at a rate only slightly better than the fourth­ hatched chicks from four-egg clutches (4 of 7, or 57%). The number of eggs that hatched also heavily influenced the fledging success of chicks relative to hatching order (Table 3-10). If two eggs hatched in three­ or four-egg clutches, the two chicks had an approximately equal chance of fledging. If more than two eggs hatched, later-hatched chicks

145 generally had a lower chance of fledging than earlier-hatched chicks; i.e., brood reduction was likely to occur.

Consideration of Hatching Asynchrony Hypotheses

I spent a small amount of time measuring feeding rates at

Laysan Finch nests, but the samples are too small to be useful. I cannot evaluate the hypothesis that asynchrony reduces peak food demand by chicks, because I do not know when the peak demand occurs for these birds. It may occur at the point of maximum chick growth (the inflection point in the growth curve), or it may occur at an older age, when the growth rate is still high but the maintenance requirements are also high due to the chick's larger mass. In Snow Buntings and Lapland

Longspurs, Hussel (1972) thought that maximum demand occurred when the chicks were 6-10 days old, although it is unclear whether fledglings need more or less food than nestlings (Husby 1986).

The sibling rivalry hypothesis does not necessarily exclude the possibility of brood reduction. Although I did not study sibling rivalry, this hypothesis is not inconsistent with my data.

I did not do detailed studies on predation on parents at the nest or on survival of parents between nesting attempts. However, many banded Laysan Finches have reached an age of at least six years old, and at least some parents have good survival between nesting seasons (Morin and Conant unpubl. data). There is little or no predation on after­ hatch-year (AHY) finches and no evidence that incubating females suffered any predation at the nest (pers. obs.), so there is little or no selection pressure against hatching asynchrony. Therefore, Magrath's

(1988) hypothesis can be discounted.

146 I cannot evaluate the hypothesis that asynchrony ensures equal parental investment in progeny of each sex. The sex ratio of HY birds is approximately 50/50 (Conant pers. corom.), suggesting that the unbalanced sex ratio which seems to occur in older age groups comes about after the first year. However, the primary sex ratio is unknown.

Slagsvold (1990) proposed that hatching asynchrony increased with an increase in size dimorphism between the sexes; Laysan Finches are slightly size dimorphic in some measurements (Conant 1988).

I have previously presented evidence (see Chapter I) that intraspecific interference competition (via egg predation) occurs in

Laysan Finches, and that food resources on Laysan apparently fluctuate between years in relation to weather. This suggested that the nest failure hypothesis, as well as the brood reduction hypothesis, might explain the observed hatching asynchrony. The brood reduction and nest failure hypotheses predict that incubation may start with the first egg laid, while the physiological constraint hypothesis predicts that incubation should begin with the penultimate egg. It is unclear which of these predictions is fulfilled; finches often begin incubation with the first egg (as evidenced by the observed hatching spreads for each clutch size, Table 3-1), but sometimes they begin with the second egg.

The brood reduction hypothesis predicts that a small final egg would facilitate the preferential starvation of the smallest chick, if necessary; my data show that the final-laid egg in Laysan Finch clutches of three or more eggs is relatively small, but so is the first-laid egg

(Figure 3-1). This hypothesis also predicts a high growth rate so that the size hierarchy within the asynchronously hatched brood may be

147

._------maintained. Relative to temperate passerines, Laysan Finches have a low

growth rate that may vary according to hatching order (Tables 3-4 and 3­

5). Although the growth rate is slow in relation to many temperate

passerines, it is probably optimal for the ambient environmental conditions, so the latter prediction of high growth rate may be met in this ecological context. Fat is stored in some chicks, and preferentially in the chicks earliest in the hatching order (Table 3-8);

fledging success in chicks was positively associated with the amount of

furcular fat. The brood reduction hypothesis predicts that at least sometimes the later-hatched chicks will have a lower fledging rate.

Preferential mortality of later-hatched chicks occurred in clutches where more than two eggs hatched (Table 3-10); this appears to support the brood reduction hypothesis. Whether this is adaptive brood reduction, or simply an inevitable result of other selection pressures, is unclear.

slagsvold et ale (1984) proposed that if an open nester (such as the Laysan Finch) showed asynchronous hatching primarily as a response to nest failure, then a brood survival strategy of relatively larger final eggs should be adopted. This strategy is not confirmed by my study (Figure 3-1). According to the nest failure hypothesis, hatching asynchrony should be favored when mortality on eggs is equal to or greater than mortality on chicks. In the Laysan Finch's case, the majority of the mortality occurs in the egg stage, presumably due to avian predation (see Chapter I). Hatching asynchrony has been viewed as an adaptation to get eggs hatched and out of the nest as quickly as possible. Although not explicitly stated as part of the nest failure

148 hypothesis, asynchrony may also be advantageous because the female is continuously guarding the eggs by covering them with her body. Laysan

Finches often begins incubation with the first egg, and may even be

found in the nest cup the day before the first egg is laid.

The insurance hypothesis is also partly supported by my results; one of its predictions is a high rate of addled or infertile eggs, or a

high rate of partial clutch loss. All Laysan Finch clutch sizes had

high rates of partial clutch loss (presumably eggs disappeared primarily because of intraspecific predation), although the percent of eggs that disappeared generally decreased as clutch size increased (see Chapter

I). Also, four-egg clutches had unusually high rates of addled or

infertile eggs. Only one four-egg clutch fledged four chicks (Table 3­

12), but a very high percent fledged three chicks (Table 3-11).

The physiological constraint hypothesis is difficult to evaluate,

since the only specific prediction is that incubation should begin with the penultimate egg, which does not occur regularly in Laysan Finches

(Table 3-1). Laysan Finches do not appear to be totally constrained by their incubation physiology. They apparently incubate while they are

still ovulating, even though it is suspected for birds in general that

incubation is induced by prolactin, and ovulation is halted by prolactin

(Drent 1975, Mead and Morton 1985). The physiological constraint

hypothesis appears to be disproved in this particular instance, or other

overriding selection pressures are also operating to promote Laysan

Finches to begin incubation on the first egg.

Clearly, predictions of the brood reduction hypothesis are

generally well supported by my data. The insurance and nest failure

149 hypotheses may also partially explain hatching asynchrony in this species. However, the strongest predictions of these hypotheses involve comparisons of synchronous and asynchronous broods in the same environment, and my data set precludes such comparisons. Although my data do not address the sibling rivalry and peak demand reduction hypotheses, either of these mechanisms may contribute to the observed hatching asynchrony. Mead and Morton (1985) suggested that ovulation physiology forces asynchrony, although it may also be modified by selection on other aspects of the reproductive strategy. This view is consistent with that of several authors (Mock 1984, Murphy and Haukioja

1986, slagsvold 1990) who believe that multiple causation for asynchronous hatching cannot be discounted.

SUMMARY

Laysan Finch chicks were observed to hatch asynchronously.

Clutches took from one to five days to complete hatching. There were few clutches in which every egg hatched or every hatchling fledged.

At age 9/10 days old, tarsus length, beak depth, beak width, beak length, wing length, and weight varied significantly between chicks that ultimately fledged and those that did not fledge; sternum length did not vary significantly. Sternum length is probably the most reliable age predictor for Laysan Finch chicks. At age 9/10 days old, body weights and sometimes beak lengths varied significantly among chicks of different hatching order in clutches where more than two eggs hatched.

In clutches where only two eggs hatched, none of the seven measurements mentioned above varied significantly.

150 The amount of fat accumulation at banding age was independent of chick fresh egg weight. Chicks that hatched earlier in the hatch sequence had significantly more furcular fat than later-hatched chicks.

Chick fat at banding age (days 11 to 14) was significantly associated with fledging success.

Average size eggs had the highest probability of producing fledglings. In clutches where only two eggs hatched, fledging success was approximately equal for both chicks. In clutches where more than two eggs hatched, chicks that hatched earlier in the hatch order had higher fledging success than later-hatched chicks. Where only one egg hatched in a nest, the chick was less likely to fledge than a third­ hatched chick in a nest with other chicks. overall, four-egg clutches produced more fledglings per nest than three-egg clutches, even though the latter was the mode. Although they were more productive, four-egg clutches included twice as many nests with addled or infertile eggs as did three-egg clutches. The modal number of fledglings per nest (in nests where at least one chick fledged) shifted from three in 1986 to two in 1987, and corresponded with a drastic reduction of rainfall between 1986 and 1987.

The brood reduction, nest failure, and insurance hypotheses are each partially supportable as explanations for Laysan Finch hatching asynchrony. The data do not permit assessment of other commonly hypothesized explanations for this phenomenon. Hatching asynchrony in this species may be maintained by a combination of influences involving a number of hypotheses.

151 LITERATURE CITED

Amundsen, T. and J. N. Stokland. 1988. Adaptive significance of

asynchronous hatching in the Shag: a test of the brood reduction

hypothesis. Jour. of Anim. Ecol. 57: 329-344.

Baldwin, S. P., H. C. Oberholser, and L.G. Worley. 1931. Measurements

of birds. Sci. Publ. of the Cleveland Mus. of Nat. Hist. Vol. II.

Cleveland. 165 pp.

Bancroft, G. T. 1984. Growth and sexual dimorphism of the boat-tailed

grackle. Condor 86: 423-432.

Blem, C. R. 1990. Avian energy storage. Chapter 2. In Current

ornithology. Vol. 7. D. M. Power (ed.). Plenum Press, New York and

London. 388 pp.

Bradley, D.W., R.E. Landry, and C.T. Collins. 1984. The use of

jackknife confidence intervals with the Richards curve for describing

avian growth patterns. Bull. Southern Calif. Acad. Sci. 83: 133-147.

Boyce, M. S. and C. M. Perrins. 1987. Optimizing Great Tit clutch size

in a fluctuating environment. Ecology 68: 142-153.

Clark, A. B. and D. S. Wilson. 1981. Avian breeding adaptations:

hatching asynchrony, brood reduction, and nest failure. Quart. Rev.

of BioI. 56: 253-277.

Clark, A. B. and D. S. Wilson. 1985. The onset of incubation in birds.

Amer. Nat. 125: 603-611.

Conant, S. 1988. Geographic variation in the Laysan Finch (Telespyza

cantans). Evol. Ecol. 2: 270-282.

152 Drent, R. 1975. Incubation. pp. 333-420. In Avian biology. Vol. V.

D. S. Farner, J. R. King, and K. C. Parkes (eds.). Academic Press,

Inc., New York.

Edwards, T. C., Jr. and M. W. Collopy. 19B3. Obligate and facultative

brood reduction in eagles: an examination of factors that influence

fratricide. Auk 100: 630-635.

Ely, C. A. and R. B. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll Res. Bull., No. 171.

Smithsonian Institution, Washington, D.C. 361 pp.

Fujioka, M. 19B5. Food delivery and sibling competition in

experimentally even-aged broods of the Cattle Egret. Behav. Ecol.

Sociobiol. 17: 67-74.

Hahn, D. C. 1981. Asynchronous hatching in the Laughing Gull: cutting

losses and reducing rivalry. Anim. Behav. 29: 421-427.

Howe, H. F. 197B. Initial investment, clutch size, and brood reduction

in the common grackle (Quiscalus guiscula L.). Ecology 59: 1109­

1122.

Husby, M. 1986. On the adaptive value of brood reduction in birds:

experiments with the magpie Pica pica. J. Anim. Ecol. 55: 75-B3.

Hussel, D. J. T. 1972. Factors affecting clutch size in arctic

passerines. Ecol. Monogr. 42: 317-364.

Johnson, D. H., A. B. Sargeant, and S. H. Allen. 1975. Fitting

Richards' curve to data of diverse origins. Growth 39: 315-330.

Kepler, C.B. 1969. Breeding biology of the Blue-footed Booby Sula

dactylatra personata on Green Island, Kure Atoll. Pub. of the

Nuttall Ornith. Club, No. B., Cambridge, MA. 97 pp.

153 Lack, D. 1954. The natural regulation of animal numbers. Oxford at

the Clarendon Press, England. 343 pp.

Lamoureux, C. H. 1963. The flora and vegetation of Laysan Island.

Atoll Res. Bull., No. 97. Smithsonian Institution, Washington, D.C.

14 pp.

Magrath, R. D. 1988. Hatching asynchrony in altricial birds: nest

failure and adult survival. Am. Nat. 131: 893-900.

Mead, P. S. and M. L. Morton. 1985. Hatching asynchrony in the

Mountain White-crowned Sparrow (Zonotrichia leucophrys oriantha): a

selected or incidental trait? Auk 102(4): 781-792.

Mock, D. W. 1984. Infanticide, siblicide, and avian nestling

mortality. In G. Hausfater and S. B. Hrdy (eds.). Infanticide:

comparative and evolutionary perspectives. Aldine, New York.

Mock, D. W. and G. A. Parker. 1986. Advantages and disadvantages of

egret and heron brood reduction. Evolution 40: 459-470.

Mock, D. W. and B. J. Ploger. 1987. Parental manipulation of optimal

hatch asynchrony: an experimental study. Anim. Behav. 35: 150-160.

Morin, M. P. In prep. The breeding biology of an endangered Hawaiian

honeycreeper, the Laysan Finch (Telespiza cantans).

Morton, E. S. 1973. On the evolutionary advantages and disadvantages

of fruit eating in tropical birds. Amer. Nat. 107: 8-22.

Murphy, M. T. 1986. Body size and condition, timing of breeding, and

aspects of egg production in Eastern Kingbirds. Auk 103: 465-476.

Murphy, E. C. and E. Haukioja. 1986. Clutch size in nidicolous birds.

Chapter 4 In Current ornithology. Vol. 4. R.F. Johnston, Ed.

Plenum Press, New York. 324 pp.

154 Newman, A. L. 1988. Mapping and monitoring vegetation change on Laysan

Island. M.A. thesis, Geography Dept., Univ. of Hawaii, Honolulu.

234 pp.

O'Connor, R. J. 1978a. Growth strategies in nestling passerines.

Living Bird 16: 209-238.

O'Connor, R. J. 1978b. structure in avian growth patterns: a

multivariate study of passerine development. J. Zool. 185: 147-172.

O'Connor, R. J. 1984. The growth and development of birds. John Wiley

and Sons, Ltd., England. 315 pp.

Parsons, J. 1975. Asynchronous hatching and chick mortality in the

Herring Gull Larus argentatus. Ibis 117: 517-520.

Richards, F. J. 1959. A flexible growth function for empirical use.

Jour. of Exper. Bot. 10: 290-300.

Richter, W. 1984. Nestling surviva~ and growth in the Yellow-headed

Blackbird, Xanthocephalus xanthocephalus. Ecology 65: 597-608.

Ricklefs, R. E. 1976. Growth rates of birds in the humid New World

tropics. Ibis 118: 179-207.

Ricklefs, R. E. 1983. Avian postnatal development. In Avian biology

VII. D.S. Farner, J.R. King, and K. Parkes (eds.). Academic Press,

Inc., 111 Fifth Ave, New York. 542 pp.

SAS Institute Inc. 1988. SAS/STAT User's Guide, Release 6.03 Edition.

Cary, NC; SAS Institute Inc. 1028 pp.

Schifferli, L. 1973. The effect of egg weight on the subsequent growth

of nestling Great Tits Parus major. Ibis 115: 549-558.

155 slagsvold, T., J. Sandvik, G. Rofstad, O. Lorensen, and M. Husby. 1984.

On the adaptive value of intraclutch egg-size variation in birds.

Auk 101: 685-697.

Slagsvold, T. 1988. Hatching asynchrony in altricial birds: nest

failure and adult survival. Amer. Natural. 131: 893-900.

Slagsvold, T. 1990. Fisher's sex ratio theory may explain hatching

patterns in birds. Evolution 44: 1009-1017. smith, K. G. 1988. Clutch-size dependent asynchronous hatching and

brood reduction in Junco hyemalis. Auk 105: 200-203.

Zach, R., Y. Liner, G. L. Rigby, and K. R. Mayoh. 1984. Growth curve

analysis of birds: the Richards model and procedural problems. Can.

J. Zool. 62: 2429-2435.

156 TABLE 3-1. Frequency of clutches in different hatching spread categories (in days). Additional clutches in which only one egg hatched appear in parentheses.

Initial HATCHING SPREAD (days) clutch size 1 234 5

two eggs o (3) o o o o three eggs 11 (9) 23 12 2 o four eggs o (7) 10 9 2 o five eggs o (2) o o 1 1

157 TABLE 3-2. Mean weight difference (g) between the last-hatched nestling (1 or 2 days old) and the largest nestling for various clutch sizes with different numbers of hatchlings. Initial clutches of two eggs are omitted because both eggs never hatched.

INITIAL CLUTCH SIZE FINAL NUMBER OF HATCHLINGS THREE FOUR FIVE

TWO MEAN 1.48 G 1.70 G S.D. 0.83 0.57 N 9 2

THREE MEAN 3.05 G 3.72 G S.D. 1.52 1.63 N 13 5

FOUR MEAN 4.67 G 9.50 G S.D. 0.76 N 3 1

FIVE MEAN 9.00 G S.D. N 1

158 TABLE 3-3. Means, standard deviations (S.D.) , ranges, and sample sizes (N) for measurements on fledged and non-fledged chicks, ages 9 and 10 days old combined; t-tests were used to test for significant differences.

Variable N Mean S.D. Range Probability

Tarsus Length (em)

Fledged 182 1.979 0.157 1.421-2.287 P<.Ol Non-fledged 16 1.830 0.249 1.360-2.284

Beak Depth (em)

Fledged 182 0.627 0.042 0.510-0.740 P<.002 Non-fledged 16 0.578 0.059 0.500-0.684

Beak Width (em)

Fledged 182 0.553 0.036 0.452-0.640 P<.002 Non-fledged 16 0.523 0.039 0.470-0.615

Beak Length (em)

Fledged 182 0.594 0.035 0.496-0.700 P<.002 Non-fledged 16 0.552 0.050 0.450-0.626

Wing Length (em)

Fledged 182 2.47 0.356 1.40-3.50 P<.OOl Non-fledged 16 2.07 0.465 1.20-2.75 sternum Length (em)

Fledged 182 0.834 0.094 0.563-1.110 P<.10 (NS) Non-fledged 16 0.788 0.091 0.642-0.932

Weight (g)

Fledged 182 21.20 3.300 8.50-29.0 P<.OO1 Non-fledged 16 15.05 5.523 6.50-24.5

159 TABLE 3-4. Summary of growth parameters estimated for the Richards' model from daily weights of fledged chicks of different hatching orders from three-egg clutches. 95% upper and lower jackknife confidence intervals (C.I.) appear in parentheses. Singles are chicks from clutches where only one egg hatched.

HATCHING ORDER

FIRST SECOND THIRD SINGLE

Asymptote (A) in grams 32.8 30.6 32.7 34.7

Upper C.L (35.7) (35.9) (44.9) (35.1) Lower C.L (30.1 ) (26.0) (23.8) (34.3)

Weighted mean growth rate (R) .185 .180 .165 .181 per day* Upper C. I. ( .197) (.199) (.187) ( .193) Lower C. I. (.174) ( .162) (.145) ( .170)

% of Asymptote at inflection (P) .50 .50 .47 .49

Upper C.I. (.56) ( .60) (.55) ( .58) Lower C. I. (.43 ) ( .40) ( .39) ( .40)

Days from 10% to 90% asymptote (G) 11.9 12.3 14.1 12.4

Upper C.I. (13.2) (15.7) (19.1) (15.8) Lower C.I. (10.8) ( 9.6) (10.5) ( 9.7)

Days to reach inflection (I) 7.4 7.4 7.8 8.0

Growth constant (K) per day .364 .354 .281 .341

Shape parameter (M) 1.97 1.97 1. 707 1.886

Number of chicks in sample 22 18 11 5

Number of data points 144 118 63 30

*R = KIM

160 TABLE 3-5. Summary of growth parameters estimated for the Richards' model from daily weights of fledged chicks of different hatching orders from four-egg clutches. 95% upper and lower jackknife confidence intervals (C.I.) appear in parentheses. Singles are chicks from clutches where only one egg hatched.

HATCHING ORDER

FIRST SECOND THIRD FOURTH SINGLE

Asymptote (A) in grams 32.7 34.8 28.9 34.2 34.4

Upper C. I. (38.1) (37.8) (34.4) (37.3) (36.2) Lower C.I. (28.1) (32.0) (24.3) (31.4) (32.6)

Weighted mean growth rate (R) .182 .172 .166 .190 .201 per day* Upper C. I. ( .204) (.191 ) (.191) (.226) (.252) Lower C.I. ( .162) ( .156) (.145 ) (.159) ( .161)

% of Asymptote at inflection (P) .49 .46 .51 .36 .42

Upper C. I. (.59) ( .52) (.58) (.52) (.54) Lower C.I. (.39) (.41) (.44 ) (.22) ( .30)

Days from 10% to 90% asymptote (G) 12.3 13.7 13.0 17.0 13.1

Upper C. I. (14.5) (14.8) (15.4) (36.1) (15.4) Lower C.I. (10.4) (12.7) (11. 0) (8.0) (11.1)

Days to reach inflection (I) 7.2 7.6 7.6 6.3 6.9

Growth constant (K) per day .346 .283 .348 .178 .263

Shape parameter (M) 1.90 1.64 2.09 .937 1.307

Number of chicks in sample 12 13 9 2 3

Number of data points 75 82 59 12 17

* R = KIM 161 TABLE 3-6A. Measurements of chicks at day 9/10 from three-egg clutches where two eggs hatched. Values of t are from t-tests comparing chicks that were first and second in the hatching order. Only chicks that eventually fledged are included. (Single chicks were not included in the analyses.) None of the measurements were significantly different at the 0.05 level.

HATCHING ORDER

VARIABLE FIRST SECOND SINGLE N=l1 N=10 N=5 t Value P

TARSUS MEAN 1.94 1.93 .202 0.84 1.88 LENGTH (CM) S.D. 0.11 0.22 0.13

BEAK MEAN 0.62 0.62 .097 0.92 0.61 DEPTH (CM) S.D. 0.04 0.06 0.01

BEAK MEAN 0.54 0.54 -.185 0.86 0.54 WIDTH (CM) S.D. 0.03 0.04 0.04

BEAK MEAN 0.59 0.59 -.263 0.80 0.57 LENGTH (CM) S.D. 0.02 0.05 0.04

WING MEAN 2.29 2.38 -.567 0.58 2.17 LENGTH (CM) S.D. 0.23 0.47 0.28

STERNUM MEAN 0.81 0.84 -.668 0.52 0.80 LENGTH (CM) S.D. 0.06 0.12 0.09

WEIGHT MEAN 20.94 20.70 .132 0.90 20.10 (G) S.D. 2.43 5.17 2.43

162 TABLE 3-6B. Measurements of chicks at day 9/10 from three-egg clutches where three eggs hatched. P values are from a one-way Analysis of Variance (ANOVA) with hatching order as treatment. Only chicks that eventually fledged are included. (Single chicks were not included in the analyses.) Only weights* were significantly different at the 0.05 level.

HATCHING ORDER

VARIABLE FIRST SECOND THIRD SINGLE N=ll N=8 N=ll N=5 P

TARSUS MEAN 2.02 2.06 1.92 0.13 1.88 LENGTH (CM) S.D. 0.11 0.13 0.18 0.13

BEAK MEAN 0.64 0.65 0.61 0.15 0.61 DEPTH (CM) S.D. 0.05 0.04 0.04 0.01

BEAK MEAN 0.57 0.59 0.54 0.08 0.54 WIDTH (CM) S.D. 0.05 0.02 0.03 0.04

BEAK MEAN 0.61 0.61 0.59 0.66 0.57 LENGTH (CM) S.D. 0.04 0.03 0.05 0.04

WING MEAN 2.56 2.63 2.43 0.55 2.17 LENGTH (CM) S.D. 0.44 0.33 0.38 0.28

STERNUM MEAN 0.88 0.88 0.84 0.57 0.80 LENGTH (CM) S.D. 0.11 0.09 0.10 0.09

WEIGHT MEAN 22.36 22.35 19.05 0.03* 20.10 (G) S.D. 2.70 2.50 3.87 2.43

163 TABLE 3-7A. Measurements of chicks at day 9/10 from four-egg clutches where two eggs hatched. Values of t are from t-tests comparing chicks that were first and second in the hatching order. Only chicks that eventually fledged are included. (Single chicks were not included in the analyses. ) None of the measurements were significantly different at the 0.05 level.

HATCHING ORDER

VARIABLE FIRST SECOND SINGLE N=5 N=4 N=3 t Value P

TARSUS MEAN 2.05 1.92 1.338 0.22 1.90 LENGTH (CM) S.D. 0.16 0.12 0.12

BEAK MEAN 0.64 0.63 .528 0.61 0.63 DEPTH (CM) S.D. 0.04 0.04 0.03

BEAK MEAN 0.55 0.56 -.118 0.91 0.53 WIDTH (CM) S.D. 0.04 0.03 0.01

BEAK MEAN 0.59 0.58 .326 0.75 0.60 LENGTH (CM) S.D. 0.04 0.04 0.05

WING MEAN 2.54 2.39 .860 0.42 2.35 LENGTH (CM) S.D. 0.30 0.20 0.43

STERNUM MEAN 0.88 0.84 .512 0.62 0.89 LENGTH (CM) S.D. 0.10 0.11 0.14

WEIGHT MEAN 22.40 19.75 1.943 0.09 21.60 (G) S.D. 2.33 1.56 2.65

164 TABLE 3-7B. Measurements of chicks at day 9/10 from four-egg clutches where three eggs hatched. P values are from a one-way Analysis of Variance (ANOVA) with hatching order as treatment. Only chicks that eventually fledged are included. (Single chicks were not included in the analyses.) Only beak lengths* were significantly different at the 0.05 level, although chick weights were very close to significance.

HATCHING ORDER

VARIABLE FIRST SECOND THIRD SINGLE N=5 N=5 N=6 N=3 P

TARSUS MEAN 2.13 2.02 1.94 0.30 1.90 LENGTH (CM) S.D. 0.12 0.24 0.20 0.12

BEAK MEAN 0.64 0.62 0.60 0.45 0.63 DEPTH (CM) S.D. 0.03 0.06 0.06 0.03

BEAK MEAN 0.57 0.56 0.54 0.66 0.53 WIDTH (CM) S.D. 0.02 0.07 0.05 0.01

BEAK MEAN 0.62 0.59 0.57 0.04* 0.60 LENGTH (CM) S.D. 0.03 0.03 0.04 0.05

WING MEAN 2.72 2.47 2.30 0.28 2.35 LENGTH (CM) S.D. 0.36 0.46 0.41 0.43

STERNUM MEAN 0.87 0.84 0.82 0.86 0.89 LENGTH (CM) S.D. 0.13 0.21 0.13 0.14

WEIGHT MEAN 23.60 22.30 19.00 0.06 21.60 (G) S.D. 2.33 3.72 2.86 2.65

165 TABLE 3-7C. Measurements of chicks at day 9/10 from four-egg clutches where four eggs hatched. P values are from a one-way Analysis of Variance (ANOVA) with hatching order as treatment. Only chicks that eventually fledged are included. (Single chicks were not included in the analyses.) Beak lengths* and chick weights* were significantly different at the 0.05 level.

HATCHING ORDER

VARIABLE FIRST SECOND THIRD FOURTH N=2 N=3 N=3 N=l P

TARSUS MEAN 1.95 2.09 1.87 1.89 .33 LENGTH (CM) S.D. 0.23 0.12 0.08

BEAK MEAN 0.62 0.64 0.59 0.58 0.11 DEPTH (CM) S.D. 0.03 0.03 0.01

BEAK MEAN 0.56 0.58 0.54 0.53 0.17 WIDTH (CM) S.D. 0.03 0.02 0.01

BEAK MEAN 0.59 0.62 0.58 0.57 0.04* LENGTH (CM) S.D. 0.01 0.02 0.01

WING MEAN 2.45 2.70 2.22 2.10 0.07 LENGTH (CM) S.D. 0.35 0.10 0.13

STERNUM MEAN 0.88 0.91 0.75 0.74 0.10 LENGTH (CM) S.D. 0.07 0.08 0.05

WEIGHT MEAN 21.00 22.00 17.50 17.00 0.04* (G) S.D. 2.83 0.50 1.00

166 TABLE 3-8. Number of chicks with each fat status at banding age (days 10 to 15) for chicks of known hatching order. "Single" chicks (from nests where only one egg hatched) were not included.

FAT STATUS

ZERO ONE TWO TOTAL

HATCHING ORDER

FIRST 2 10 38 50

SECOND 2 12 25 39

THIRD 8 10 15 33

FOURTH 2 2 1 5

127

167 TABLE 3-9. Frequency of hatching categories for different clutch sizes, 1986 to 1988 combined. Only known-size clutches were used.

NONE HATCH SOME HATCH ALL HATCH

CLUTCH SIZE TOTAL

ONE 1 0 0 1

TWO 15 3 0 18

THREE 43 33 24 100

FOUR 12 22 6 40

FIVE 1 2 (OR 3) 3 (OR 2) 6

168 TABLE 3-10. Percent of hatchlings that fledged from known-size clutches relative to their hatching order, 1986 to 1988 data combined. N indicates the number of nests.

Initial Clutch Size and Hatching Order

2-EGG 3-EGG 4-EGG 5-EGG 1st 2nd 1st 2nd 3rd 1st 2nd 3rd 4th 1st 2nd 3rd 4th 5th

Number of Hatchlings

1 67% 71% 67% 50% N=3 N=7 N=6 N=2

2 58% 58% 67% 57% N=24 N=9

3 83% 74% 61% 100% 100% 86% N=23 N=7

4 75% 100% 100% 50% 100% 100% 100% 0% -- N=4 N=l

5 100% 100% 100% 100% 0% N=l

169 TABLE 3-11. Frequency of occurrence of nests with various numbers of fledglings for clutches of known size, for nests in 1986 to 1988 combined. Percent of nests for each clutch size is shown in parentheses.

CLUTCH SIZE

NUMBER OF FLEDGLINGS

PER NEST 1 2 3 4 5

ZERO 1 (100%) 16 (89%) 58 (58%) 16 (40%) 2 (33%)

ONE 0 2 (11%) 12 (12%) 6 (15%) 2 (33%)

TWO 0 18 (18%) 6 (15%) 0

THREE 12 (12%) 11 (28%) 1 (17%)

FOUR 1 (2%) 1 (17%)

FIVE 0

TOTALS 1 18 100 40 6

170

------~. -_.-._--- .-. ---- TABLE 3-12. Frequency of occurrence of nests with various numbers of fledglings for clutches of known size for nests in 1986. Percent of nests for each clutch size is shown in parentheses.

CLUTCH SIZE

NUMBER OF FLEDGLINGS

PER NEST 1 2 3 4 5

ZERO 1 (100%) 8 (89%) 31 (70%) 7 (39%) 1 (33%)

ONE 0 1 (11%) 3 (7%) 3 (17%) 0

TWO 0 2 (5%) 2 (11%) 0

THREE 8 (18%) 5 (28%) 1 (33%)

FOUR 1 (5%) 1 (33%)

FIVE 0

TOTAL NESTS 1 9 44 18 3

171 TABLE 3-13. Frequency of occurrence of nests with various numbers of fledglings for clutches of known size for nests in 1987. Percent of nests for each clutch size is shown in parentheses.

CLUTCH SIZE

NUMBER OF FLEDGLINGS

PER NEST 1 2 3 4 5

ZERO 7 (100%) 18 (44%) 5 (41%)

ONE 0 7 (17%) 2 (17%)

TWO 0 13 (32%) 3 (25%)

THREE 3 (7%) 2 (17%)

FOUR 0

FIVE

TOTAL NESTS o 7 41 12 o

172 TABLE 3-14. Frequency of occurrence of nests with various numbers of fledglings for clutches of known size for nests in 1988. Percent of nests for each clutch size is shown in parentheses.

CLUTCH SIZE

NUMBER OF FLEDGLINGS

PER NEST 1 2 3 4 5

ZERO 1 (50%) 9 (60%) 4 (40%) 1 (33%)

ONE 1 (50%) 2 (13%) 1 (10%) 2 (66%)

TWO 0 3 (20%) 1 (10%) 0

THREE 1 (7%) 4 (40%) 0

FOUR 0 0

FIVE 0

TOTAL NESTS o 2 15 10 3

173 """'(/) 4.0 • OVERALL ~ 3.6 <> 1986 ct: D 1987 6 1988 '-oJ" 3.2 I- 2.8 :::c "w 2.4 3= 2.0 ...... ------FIRST SECOND TWO EGG CLUTCHES """' 4.0 in ~ 3.6 ct:

'-oJ" 3.2 I- 2.8 lIn HI! nIl :::c " 2.4 w 3= 2.0 ...... ~ ...... o...;._...... ;;;:;~.;..;.. _ FIRST SECOND THIRD THREE EGG CLUTCHES """' 4.0 in ~ 3.6 ct:

'-oJ" 3.2 I- 2.8 :::c " 2.4 w 3= 2.0 ...a...... ,,;,,; ....;.;.....;....;...... ;.;.~_._;..;....;..,,;._;.. _ FIRST SECOND THIRD FOURTH FOUR EGG CLUTCHES

"""'(/) 4.0 ~ 3.6 ct: I~ " 3.2 6 6 6 '-oJ II . II · r I- 2.8 :::c " 2.4 w N- I') N Q - I':J N Q - N - 0 ... 3= 2.0 FIRST SECOND THIRD FOURTH FIFTH FIVE EGG CLUTCHES FIGURE 3-1 Mean fresh egg weights (grams±S.D.) by laying order for all clutch sizes by year. N indicates number of eggs.

174

------3.0

2.5 I ...... IIIlIII ~ (..) ~ 2.0 ::I: l- e.!) I I I I Z lLJ 1.5 -I en en::J I I ~ -c 1.0 l- I I I I 0.5

0.0 --~...... --a...... -a.--a.------"""''''''''''''''''--f.~''''''' o 1 2 3 4 5 678 91011121314151617 >30 AGE (DAYS)

FIGURE 3-2

Mean (±S.D.) tarsus lengths (em) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

175 1.3

1.2 1. 1 I ...... 1.0 ~ o "-J 0.9 ::c J-- 0.. 0.8 w I 9: I 0 0.7 ~ I I -c w 0.6 I I I I I OJ 0.5 I I I I 0.4 I I I 0.3

0.2 a....ll...... l --L --L --f~--.II o 1 2 3 4 5 6 7 8 9 1011121314151617 >30 AGE (DAYS)

FIGURE 3-3

Mean (±S.D.) beak depths (em) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

176 1.3

1.2 1. 1 ,..... 1.0 :E C,) '--" 0.9 I l- Cl 0.8 I == ~ 0.7 I I -c w m 0.6 I I I I I I I I 0.5 I I 0.4 I I I I I 0.3

O. 2 ...... I...... I--&...... &...-'-...... I-..I-...... &.....L-...... I...... I~ o 1 2 345 6 7 8 9 1011121314151617 AGE (DAYS)

FIGURE 3-4

Mean (±S.D.) beak widths (em) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

177 1.3 1.2 I 1.1

~ ~ u 1.0 '-' :c 0.9 t- C) z 0.8 w I ~ I ...J

~ 0.7 -c I I I w CD 0.6 I I I 0.5 I I I 0.4 I I II 0.3 I

0.2 0 1 2 3 4 5 6 7891011121314151617 >30 AGE (DAYS)

FIGURE 3-5

Mean (±S.D.) beak lengths (em) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

178 9 8 I 7 ...... ~ o 6

-::I: I- 5 I z I "w -J 4 I I o z 3 3= I I I I 2 I I II 1 :E I I :E T 0 0 1 2 3 4 5 6 7 8 9 1011121314151617 >30 AGE (DAYS)

FIGURE 3-6

Mean (±S.D.) wing lengths (em) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

179 2.0 1.8 I 1.6 ...... ~ 1.4 .....,o ~ :c 1.2 I I-- C) z w 1.0 -J I II I ~ 0.8 :::J I I I I z a: w 0.6 I I I-- (f) I I II I 0.4

0.2

0.0 0 123456 7 8 91011121314151617 >30 AGE (DAYS)

FIGURE 3-7 Mean (±S.D.) sternum lengths (em) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

180 40 35 I 30

25 "~

'-J .-" 20 :c "w ~ 15 II 10 I I 5 I I I 0 0 1 234 5 67891011121314151617 >30 AGE (DAYS)

FIGURE 3-8

Mean (±S.D.) weights (gm) plotted against age (days) for chicks which later fledged. See Appendix A for sample sizes.

181

------35 o o ...._...---_ ...... -- 30 ------

25 o ...... ~ ....., 20 D " D I-::r:

15 D "w 6 3= 6 10 6 D o FIRST HATCHED CHICK (N=22) c D SECOND HATCHED CHICK (N=18) 5 D 6 THIRD HATCHED CHICK (N=11)

O .....------,r-----or-----..,..-----r-----, o 5 10 15 20 25 AGE (DAYS)

FIGURE 3-9

Richards curves for first-. second-, and third-hatched chicks from three-egg clutches where two or three eggs hatched. The solid line represents the first chicks, the long dash line second chicks. and the short dash line third chicks. N is the number of chicks.

182 APPENDIX A. Means, standard deviations, and sample sizes for measurements taken at a given age (in days) for chicks that ultimately fledged. All linear measurements are in centimeters; weight is in grams. Age > 30 days indicates HY (Hatch Year) birds that have fledged but whose exact age is unknown.

MEASUREMENT

AGE TARSUS BEAK BEAK BEAK WING STERNUM WEIGHT LENGTH DEPTH WIDTH LENGTH LENGTH LENGTH (DAYS) (CM) (CM) (CM) (CM) (CM) (CM) (G)

ONE MEAN 0.70 0.36 0.35 0.31 0.65 0.44 2.92 S.D. 0.04 0.02 0.02 0.02 0.06 0.04 0.54 N 30 14 12 30 30 30 30

TWO MEAN 0.79 0.38 0.36 0.34 0.73 0.47 3.78 S.D. 0.06 0.03 0.03 0.03 0.05 0.05 0.89 N 65 59 58 65 65 65 65

THREE MEAN 0.89 0.41 0.38 0.37 0.82 0.51 5.46 S.D. 0.07 0.03 0.03 0.03 0.06 0.05 1.12 N 100 99 ... .99 100 100 100 100

FOUR MEAN 1.02 0.44 0.41 0.40 0.95 0.55 7.40 S.D. 0.10 0.03 0.04 0.03 0.10 0.06 1. 73 N 77 77 77 77 77 77 77

FIVE MEAN 1.20 0.47 0.43 0.44 1.12 0.60 9.56 S.D. 0.11 0.03 0.04 0.03 0.13 0.07 1.72 N 101 101 101 101 101 101 101

SIX MEAN 1.36 0.51 0.47 0.48 1.35 0.64 12.07 S.D. 0.15 0.04 0.04 0.03 0.20 0.08 2.37 N 78 78 78 78 78 78 78

SEVEN MEAN 1.58 0.55 0.49 0.51 1.63 0.70 14.62 S.D. 0.12 0.04 0.03 0.03 0.20 0.08 2.29 N 98 98 98 98 98 98 98

EIGHT MEAN 1. 74 0.58 0.52 0.54 1.95 0.75 17.19 S.D. 0.17 0.04 0.04 0.04 0.30 0.09 3.09 N 79 80 80 80 80 80 80

183 APPENDIX A (Continued)

TARSUS BEAK DEP. BEAK WID. BEAK LEN. WING STERNUM WEIGHT NINE MEAN 1.92 0.61 0.54 0.58 2.31 0.81 20.22 S.D. 0.12 0.04 0.03 0.03 0.25 0.08 2.61 N 100 100 100 100 100 100 100

TEN MEAN 2.05 0.64 0.57 0.61 2.65 0.87 22.39 S.D. 0.17 0.04 0.03 0.04 0.38 0.10 3.66 N 82 82 82 82 82 82 82

ELEVEN MEAN 2.20 0.67 0.59 0.64 3.03 0.91 24.36 S.D. 0.14 0.04 0.03 0.04 0.32 0.10 3.53 N 91 91 91 91 91 91 91

TWELVE MEAN 2.28 0.69 0.60 0.67 3.39 0.96 26.42 S.D. 0.14 0.04 0.03 0.04 0.40 0.12 3.18 N 63 63 63 63 63 63 63

THIRTEEN MEAN 2.38 0.71 0.63 0.70 3.75 1.00 28.18 S.D. 0.13 0.04 0.03 0.03 0.33 0.11 3.16 N 63 63 63 63 63 63 63

FOURTEEN MEAN 2.44 0.73 0.64 0.73 4.05 1.06 29.00 S.D. 0.12 0.04 0.04 0.04 0.38 0.10 3.36 N 34 34 34 34 34 34 34

FIFTEEN MEAN 2.48 0.77 0.67 0.77 4.39 1.15 30.45 S.D. 0.13 0.04 0.03 0.03 0.30 0.12 2.37 N 17 17 17 17 17 17 17

SIXTEEN MEAN 2.54 0.81 0.71 0.80 4.95 1.32 33.17 S.D. 0.06 0.01 0.02 0.01 0.13 0.02 2.02 N 3 3 3 3 3 3 3

SEVENTEE MEAN 2.51 0.82 0.71 0.80 5.03 1.24 33.13 S.D. 0.14 0.04 0.04 0.05 0.36 0.10 3.11 N 6 6 6 6 6 6 6

> THIRTY MEAN 2.56 1.16 0.80 1.21 7.92 1. 78 34.49 S.D. 0.08 0.05 0.03 0.05 0.22 0.09 2.71 N 240 240 239 240 240 229 237

184 APPENDIX B. Means, standard deviations, and sample sizes for measurements taken at a given age (in days) for chicks that died prior to fledging. All linear measurements are in centimeters; weight is in grams.

MEASUREMENT

AGE TARSUS BEAK BEAK BEAK WING STERNUM WEIGHT LENGTH DEPTH WIDTH LENGTH LENGTH LENGTH (DAYS) (CM) (eM) (eM) (CM) (CM) (CM) (G)

ONE MEAN 0.71 0.35 0.34 0.30 0.66 0.46 2.82 S.D. 0.04 0.02 0.03 0.02 0.06 0.04 0.50 N 17 11 11 17 17 16 17

TWO MEAN 0.77 0.37 0.36 0.34 0.71 0.48 3.48 S.D. 0.05 0.02 0.02 0.03 0.04 0.05 0.72 N 29 23 23 28 28 28 28

THREE MEAN 0.85 0.39 0.37 0.37 0.77 0.50 4.40 S.D. 0.08 0.03 0.03 0.03 0.08 0.05 1.44 N 21 20 20 21 21 21 21

FOUR MEAN 1.00 0.42 0.39 0.39 0.88 0.55 6.12 S.D. 0.07 0.03 0.03 0.02 0.06 0.06 1.29 N 20 20 20 20 20 20 20

FIVE MEAN 1.13 0.45 0.42 0.44 1.02 0.58 7.35 S.D. 0.12 0.04 0.03 0.04 0.14 0.06 1.95 N 14 14 14 14 14 13 14

SIX MEAN 1.30 0.49 0.44 0.47 1.24 0.63 9.50 S.D. 0.11 0.04 0.02 0.03 0.16 0.05 2.70 N 12 12 12 12 12 12 12

SEVEN MEAN 1.45 0.51 0.46 0.50 1.37 0.68 10.23 S.D. 0.19 0.05 0.04 0.04 0.30 0.09 3.32 N 9 9 9 9 9 9 9

EIGHT MEAN 1.62 0.54 0.50 0.51 1.68 0.70 13.39 S.D. 0.22 0.06 0.04 0.05 0.29 0.06 4.86 N 11 11 11 11 11 11 11

185

-_._-----_._--- APPENDIX B (Continued)

TARSUS BEAK DEP. BEAK WID. BEAK LEN. WING STERNUM WEIGHT NINE MEAN 1. 78 0.57 0.51 0.54 1.93 0.75 14.13 S.D. 0.23 0.05 0.03 0.05 0.43 0.08 4.32 N 8 8 8 8 8 8 8

TEN MEAN 1.88 0.59 0.54 0.57 2.20 0.82 15.98 S.D. 0.27 0.06 0.04 0.05 0.48 0.09 6.69 N 8 8 8 8 8 8 8

ELEVEN MEAN 2.07 0.61 0.55 0.61 2.61 0.86 18.44 S.D. 0.14 0.05 0.03 0.03 0.44 0.12 3.71 N 7 7 7 7 7 7 7

TWELVE MEAN 2.35 0.70 0.60 0.66 3.31 0.96 25.13 S.D. 0.17 0.04 0.04 0.05 0.33 0.06 6.22 N 4 4 4 4 4 4 4

THIRTEEN MEAN 2.38 0.70 0.61 0.68 3.50 0.98 28.50 S.D. 0.15 0.08 0.02 0.05 0.30 0.06 3.61 N 3 3 3 3 3 3 3

FOURTEEN MEAN 2.44 0.75 0.64 0.73 4.03 1.03 25.00 S.D. 0.18 0.05 0.06 0.03 0.13 0.07 7.55 N 3 3 3 3 3 3 3

186 CHAPTER IV

LAYSAN FINCH POPULATION ESTIMATES IN RELATION

TO THE ANNUAL BREEDING CYCLE AND OTHER VARIABLES

INTRODUCTION

The population of endangered, endemic Laysan Finches (Telespiza cantans) on Laysan Island has been censused almost yearly since 1968 using a strip transect method (Sincock and Kridler 1977). This technique was chosen because of its simplicity and ease of application, especially since the census takers are sometimes inexperienced and the time available for a count on this remote island is usually short. The purpose of the census has been to monitor population changes over time.

Within the logistic constraints of the situation, it is important to sample the population accurately, yet efficiently. The

appropriateness of the strip transect technique currently used on Laysan

has sometimes been questioned by census participants. Participants often are initially preoccupied with locating the "exact" placement of the unmarked transects, have difficulty standardizing the length of time they spend traversing each transect, or experience a variety of other

problems. This paper may help to resolve these and other questions, by

considering how well the assumptions for the strip transect method are

met, how the assumptions could be better met, and evaluating whether the

resulting population estimates are adequate for their intended purpose.

Sizes of animal populations may be estimated using a variety of methods. The most commonly used methods for estimating avian density

are spot or territory mapping (Verner and Ritter 1988), total mapping

(Desante 1981, Verner 1985), and line transect counts (Burnham et al.

187 1980, Brennan and Block 1986, Bollinger et ale 1988, Hanowski et ale

1990). Special cases of line transects include variable circular plots

(Reynolds et ale 1980, Conant et ale 1981, Bollinger et ale 1988) and strip transects (Burnham et ale 1980, Burnham and Anderson 1984, Conroy et ale 1988). Migration counts (HuSsel 1981) and indices such as call counts (Dawson 1981, Rotella and Ratti 1986) or nest counts (Bull 1981) are used to indicate population trends in situations where absolute densities are not necessary. Because few birds have populations where all individuals live in a randomly mingling flock with an equal probability of being caught, capture-recapture techniques are infrequently used in estimating population size (Nichols et a1. 1981,

Pollock et ale 1990).

Many variables are widely known to affect avian population estimates. These variables include: the time of day (Skirvin 1981), the time of year (Diehl 1981, Best and Petersen 1982), weather conditions both before and during the census (Robbins 1981), the behavior and relative detectability of the species (Burnham et ale 1980, Conant et ale 1981), differences among the observers (Kepler and Scott 1981,

Verner and Milne 1990), and the size, number, duration, and type of population samples taken (Burnham et a1. 1980, Scott and Ramsey 1981,

Verner 1985, Hanowski et ale 1990). This paper will consider some of these variables.

Strip transects are a special case of line transects (Burnham and

Anderson 1984). For line transects in general, g(x) is a detection function (probability of detecting an object at distance x from the centerline of the transect) which is equal to 1 only on the centerline

188 and declines as x increases. In contrast, strip transects assume that g(X) = 1 for all x < w, where w is one-half the width of the strip. For strip transects,

D = nj(2Lw) where D is the density estimate, n is the number of detected objects, L is the total length of the transects, and w is defined above. Verner

(1985) suggested that the major assumptions for strip transects are: (1) all birds within the strip are detected, (2) no bird moves into or out of the strip in response to the moving observer, (3) no bird is counted more than once, (4) no errors are made in determining whether a bird is within the strip, and (5) detections are independent events, especially when results from more than one transect are pooled. Verner's additional assumption (6) that all birds are correctly identified, is not violated on Laysan, since the Laysan Finch is the only passerine that still occurs there.

This paper concentrates primarily on the results of the seven population estimates made during three field seasons from 1986 to 1988 on Laysan Island, although population estimates for other years are also

'presented. Two or three censuses were made each year during 1986 to

1988; in each other year, one census or no census was taken. The stage in the finch breeding cycle was known for each census during 1986 ­

1988; it was not known in the other years. The counts were used to monitor population trends, but my primary purpose was to examine how the population estimates varied in relation to breeding cycle chronology, different vegetation types in the habitat, and individual counter performance. strip transects were used to sample the finch population

189 and derive the population estimates. During one of the post-breeding

counts, the subset of hatch-year birds was also censused in order to

estimate recruitment.

STUDY AREA

Laysan is a remote, uninhabited, coral sand island in the Hawaiian

Islands National Wildlife Refuge, located 25 0 45' Nand 1710 44' W in

the Pacific Ocean. It is about 1.7 km wide by 2.9 km long. A shallow,

hypersaline lake covers the interior of the island (USFWS 1984). During

dry spells most of the lake bed may become exposed due to evaporation.

Although at least two freshwater seeps feed into the lake, there is no

standing fresh water on the island (Ely and Clapp 1973).

The vegetation on Laysan is generally low and open, and primarily

composed of herbs, grasses, and prostrate vines. The introduced bush

Pluchea indica and a few introduced coconut palms (Cocos nucifera)

provide the only non-herbaceous vegetation inland. A few short

Tournefortia argentea trees occur along the beach at the vegetation

line; they are a minor component of the vegetation, although they are

important seabird roosting and nesting sites. The low, native shrub

Scaevola sericea occurs primarily in a narrow band on the vegetation's

outer edge, but small isolated patches are scattered about, primarily in

the bunchgrass (Eragrostis variabilis) association. Although several

vegetation zones have been described for Laysan (Lamoureux 1963, Newman

1988), the bunchgrass association and the viney association (Ipomoea­

Boerhavia-Sicyos-Tribulus) predominate (Figure 4-1). Generally, the

vegetation associations occur in concentric bands around the lake. The

190

------flora of Laysan is still largely native (Newman 1988); however, over the past century many non-native plants have become established, some of which are pestiferous.

The Laysan Finch is the only passerine still found on Laysan, although an additional three species of endemic land birds became extinct there during this century (Ely and Clapp 1973, Sincock and

Kridler 1977). Fossil evidence shows that Laysan Finches occurred on at least two of the main Hawaiian Islands, indicating a much larger natural distribution in the past (Olson and James 1982). Based on population censuses since 1968, the population estimates on Laysan have generally been close to 10,000 birds (USFWS 1984).

The only mammal which occurs on Laysan is the endangered Hawaiian

Monk Seal (Monachus schauinslandi). The endangered, endemic Laysan Duck

(~ laysanensis) also still occurs there. Laysan is an important seabird refuge which provides breeding habitat for 17 species of seabirds, including many burrow and ground nesting species. Nesting burrows occur throughout the island, in all vegetation associations, and humans frequently collapse these burrows when walking over them. A major consideration in deciding the frequency and intensity of finch censuses is the necessity to minimize seabird burrow destruction. It is also important to minimize flushing of ground nesting seabirds, and the resulting predation on their eggs by Laysan Finches. For these reasons, and also to reduce the accidental spread of pestiferous plant seeds by researchers' clothing and shoes, routine research activity (except for the censuses) was mostly confined to flagged trails or the nonvegetated shoreline.

191 METHODS

Prior to 1968, a variety of methods were used to estimate the

Laysan Finch population on Laysan Island (Brock 1951, Ely and Clapp

1973, Sincock and Kridler 1977). The current sampling technique was developed by Sincock and Kridler (1977) and has been used by the u.S.

Fish and Wildlife Service (USFWS) for each strip transect census since

1968. The vegetated area of an aerial photograph of Laysan was overlaid with a grid of 820 square sections, each 91.4 m x 91.4 m. From this grid, 120 sections were randomly chosen for the placement of numbered transects (Figure 4-2). The same 120 sections have been used during every subsequent census; however, counters have been allowed to walk their 91.4 m x 5.0 m (0.046 ha) strip transect along any boundary or diagonal for each assigned section. Due to shifting sand and the proclivity of seabirds to fly into poles and injure themselves, transects have never been permanently marked on the ground. Counters used photoreproduced aerial maps, referencing prominent landmarks or vegetation features to locate the general vicinity of each specific grid section. Approximately 3% of the total 187 vegetated hectares on Laysan was sampled during each census. Each transect was determined by the number and length of paces needed by individual counters to traverse

91.4 meters. Prior to a count, counters calibrated their paces from practice transects. From 1986 to 1988, counters were shown the width of a sample staked transect, to assist them with width estimation during the count. In an effort to standardize their estimates of transect width, counters carried light bamboo poles during the late 1960's and early 1970's. In 1988, counters carried lightweight plastic conduit

192 pipes marked off at 2.5 m (1/2 the transect width). In September 1968,

Sincock and Kridler (1977) recorded the primary vegetation association along each transect and used those data to estimate the bird population using a stratified sampling technique. They concluded that this modification produced only a minor improvement in the confidence intervals for the population estimates.

During the 1986 to 1988 censuses, two to six counters participated in each of the seven censuses. Different counters participated in from one to as many as seven censuses apiece; a total of 17 different persons were counters at least once. Research personnel on Laysan are limited during every field season due to the remoteness of the island as well as the need to protect the habitat from too much human impact. Many of the counters, of necessity, were persons involved in other types of research on Laysan, and many had never participated in a bird count. In 1988, the three counters with no previous experience in a Laysan Finch count were required to count a few "practice" transects the day before the actual count. Counters were assigned contiguous blocks of transects, so that total census time as well as seabird burrow destruction were minimized. Counters were instructed to take the censuses between the hours of 0800 and 1100, covering the transects in a slow, continuous walk, and recording each bird only once. They were instructed to cancel counts during heavy wind or any precipitation above a drizzle or light, passing shower.

The mean number of birds per transect was compared both among censuses and among individuals within each census using Kruskal-Wallis tests. Mann-Whitney tests were used to examine whether experienced

193 counters recorded significantly different rtumbers of birds per transect

than inexperienced counters within each census; a counter was classified

as experienced if he/she had participated in one or more previous Laysan

Finch censuses.

For the April 1987 census, counters were taught to identify the

five major vegetation associations, and were instructed to record the major vegetation association present on each of their transects. Based

on their records, I later assigned each transect to one of three

classifications: 1) predominantly Eragrostis grassland, 2) predominantly

viney vegetation (Ipomoea-Boerhavia-Sicyos-Tribulus), and 3) other

vegetation types. If greater than 50% of a transect was in a particular

vegetation association, I assigned the entire transect to that

vegetation type. These three classifications were later used to examine whether the sampling should be stratified according to vegetation type.

Within each census, the number of birds per transect in the three

vegetation classifications were compared by a Kruskal-Wallis test.

Since the count data were not normally distributed, an attempt was made

to normalize the data within each census: transects were "lengthened" by

summing the nearest two transects (and also the nearest four transects)

having the same vegetation classification.

During the post-breeding July 1986 census, counters were requested

to record the subset of hatch-year (HY) birds in the total number of

birds counted on transects. Hatch-year birds are easily identified in

the field by the yellow flanges of their mandibles (which are present

for at least 2 1/2 months after hatching), their vocalizations,

behavior, and fresh plumage. These data were gathered in order to

194 quantify potential recruitment, by identifying the percent of hatch-year birds in the total population estimate at that instant in time.

Each of the seven counts from 1986 to 1988 was assigned to a pre­ breeding, breeding, or post-breeding category based on my other field observations of breeding chronology (Figure 4-3). unfortunately, due to our late arrival on Laysan in 1987, the pre-breeding count (April) was done too close to the breeding count (May), since pairing and courtship had already begun. The 1987 April count should probably be considered transitional between pre-breeding and breeding status; however, even the

1987 May breeding census may have been made prior to the peak of nesting activity (see Chapter II).

The validity of the strip transect census estimates depends upon the assumption (supported by many months of field observations) that finches occur primarily in the vegetated areas of the island. The population estimates are based on the vegetated area (187 ha) rather than the total 407 ha of the island (USFWS 1984). The vegetated area on the island has remained relatively stable over the past 20 years

(Newman 1988). Population estimates for each census were computed by multiplying the average census density (birds per ha) for all 120 transects by the 187 ha. The 95% C.I. was constructed by using the population estimate ± (1.98)(S.E.)(total vegetated area/area per transect), where 1.98 was obtained from a t table for 119 d.f. (d.f. number of transects per census - 1) and S.E. was the standard error of the mean transect count within each census.

In 1987, participants in pre-1986 censuses were surveyed by mail in order to better define their census methodology and to identify

195 inconsistencies in techniques. Their comments on proposed revisions of the census technique were also solicited at that time.

Data were analyzed statistically with a personal computer software package of SAS 6.03 (SAS 1988).

RESULTS

Finch Distribution

Count data will approximate a Poisson distribution when there is a low average number of objects per sampling unit (i.e. birds are rare), and the objects occur independently (Sokal and Rohlf 1981). I compiled the transect data from all seven censuses, for a total of 837 transects.

Figure 4-4 shows the expected and the observed frequencies of birds per transect and indicates that the number of birds per transect did not have the expected Poisson distribution (Chi Square = 1538, P < .001).

There were many more transect counts with zero birds and six or more birds than expected.

The overall coefficient of dispersion (CD) for the combined data was 2.67, where CD is equal to the variance divided by the mean (Sokal and Rohlf 1981). If CD = 1, it indicates a poisson distribution, while

CD < 1 indicates a uniform or evenly spaced distribution, and CD > 1 indicates a clumped, or "contagious" distribution. For each of the seven censuses taken separately, 2 < CD < 3, indicating that there was a consistent tendency for finches to be clumped in their distribution.

There did not appear to be large seasonal changes in the CD (Table 4-2), although the April count and all three May breeding counts were the

196 least clumped. I have previously identified the April count as a questionable pre-breeding count.

Comparisons Between and Within Years

The population estimates indicated a downward trend during the three years of intense study from 1986 to 1988 (Figure 4-3), but no overall population trend over the past 22 years was evident (see

Discussion). Each of the 1986-1988 May breeding counts had the lowest population estimate within its respective year. Generally, 1986-1988 population estimates ranged from about 5,000 birds to 15,000 birds, with four of the seven estimates close to 10,000 birds (Table 4-J).

Densities (birds per transect) obtained for all transects from 1986 to 1988 were compared among the seven sampling dates. The count data were not normally distributed; hence, a Kruskal-Wallis test was made on the ranked data for the seven censuses. The number of birds per transect was significantly different among the counts (Chi Sq. = 76.248, d.f.=6, P=.OOOl). A Tukey test separated the censuses into three groups

(Figure 4-5). Except for August 1988, the May breeding censuses tended to have the lowest ranks, and the non-breeding censuses the highest ranks. Both of the 1988 population estimates were lower than any of the other five censuses from 1986 and 1987 (Table 4-1).

Densities (birds per transect) were examined within each year to compare estimates during the different breeding stages. A Kruskal­

Wallis test showed that the 1986 data (birds per transect) for the pre­ breeding (N=117), breeding (N=120), and post-breeding (N=120) censuses were significantly different (Chi Square = 10.897, d.f.= 2, P=.0043). A

197 Tukey test on the ranked transect data grouped the 1986 pre- and post­ breeding counts (February and July, respectively) as one group and the

May breeding count as a separate group. A Mann-Whitney test (again using ranked birds-per-transect for each set of 120 transects) demonstrated that the two 1987 counts, which were taken temporally close together (April and May), were not significantly different (Z = 0.851, P

= 0.400). A Mann-Whitney test showed that the two 1988 counts (a breeding count in May and a post-breeding count in August) were significantly different (Z 3.448, P = 0.001).

Estimation of Recruitment

During the July 1986 post-breeding census, all five counters kept a subtotal tally of HY birds as part of their total counts. Of the 435 birds from the transect counts, 98, or 22.5% of the birds could be identified as young of the year. The July population estimate was

14,786 ± 2,360, which gives a recruitment estimate of at least 3,327 HY birds. This is a minimum estimate of recruits. At least one counter appeared to have undercounted HY birds on the transects. Also, an earlier pulse of breeding was known to have occurred in April 1986 (see

Chapter I), even though few of those young fledged; the fledglings from that early breeding would have lost their yellow mandibular coloration by July, and could have easily been aged incorrectly by the three inexperienced counters. In addition, it is probable that a few birds were still breeding during the July count, and hence a small number of late recruits were missed.

198 Although HY birds were not tallied separately during the May and

August 1988 censuses, I know (based on other field observations; see

Chapter I), that HY birds were present. HY birds were not present during the February 1986 census nor the April and May 1987 censuses.

Comparisons Among Individual Counters

A Kruskal-Wallis test was done on the ranked densities (birds per transect; N=l20 densities per census except for February 1986 where

N=117) within each of the seven censuses to assess the effects of individual counters. Six of the seven censuses had significant differences among counters (Table 4-3). May 1987, with no significant differences, was the single census where all six counters had done at least one complete previous census. However, when each of the six censuses that used both experienced and inexperienced counters was tested separately, mean counts per transect for experienced versus inexperienced counters within each census were statistically indistinguishable in five censuses. The lone exception was the August

1988 census, which showed a significant difference in the transect counts between the single experienced counter and the single inexperienced counter (Table 4-3). The August 1988 census was the only one of the seven censuses where only two people counted (the rest had 4 to 6 counters) and it was the only census that took two days to complete.

comparisons Among Vegetation ~

The predominant vegetation type assigned to any particular transect based on the April 1987 count (see Methods) was assigned to that

199 transect for each of the seven 1986-1988 censuses. This procedure was based on the plausible assumption that counters were in approximately the same location and therefore the same vegetation association for a particular transect during each census. Most of the 120 transects fell into the grassland (N = 72) and the viney (N = 41) associations, with only 7 transects categorized as "other" vegetation types. A Kruskal­

Wallis test using data from all seven censuses showed that there was no significant difference among the counts in these three different vegetation classifications (Chi Sq = 4.723, d.f. = 2, P = 0.094).

However, when average finch counts per transect in each vegetation type were compared within each census period, none of the three May breeding censuses showed significant differences, whereas all four of the non­ breeding censuses did (Table 4-2). Both of the post-breeding counts

(July 1986 and August 1988) had the highest average number of finches per transect in those transects classified as predominantly Eragrostis grassland, the second highest average number per transect in the viney association transects, and the least number of finches in transects

categorized as "Other". Both of the pre-breeding counts (February 1986

and April 1987) had the highest average number of finches per transect

in the viney association, but the vegetation type with the second and

third highest average was not consistent.

Evaluation of Transect Length

An attempt was made to normalize the data by pairing the closest

transects of the same vegetation type, yielding 60 transects of length

182.4 m. The merged data were reexamined and determined to remain non-

200 normally distributed. Merging four adjacent transects of like vegetation type normalized the data for 1986 and 1987, but not for 1988

(when the population was low). Merging at the two levels reduced sample sizes from 120 transects to 60 or 30 transects, causing a concomitant increase in the confidence intervals for the population estimates.

Combining transects in this manner also made it impossible to examine differences among the counts of individual counters.

Survey Results from Pre-1986 Counters

Nine persons responded to the survey that asked for their recollections and comments about the methodology used for strip transect censuses before 1986. Their responses identified several problem areas.

For example, how far in front of the counter should birds be counted?

Should aural detections be recorded if they are within the transect?

How should birds moving into and out of the transect be reported? How much time should be spent on each transect? Should a pole of known length be carried so that the transect width can be more accurately estimated? Other than recording total time for the census, no counters recalled timing individual transects, but several agreed that 3 minutes per transect was probably a reasonable approximation of the "slow and easy" walk. Four of the nine respondents had either used known length poles or suggested their use for transect width estimations.

DISCUSSION

Short- and Long-term Population Trends

Laysan Finch population estimates on Laysan have shown marked fluctuations over the past 23 years, ranging from a low of approximately

201 5,000 to a high of about 20,000 (Figure 4-6). There is no obvious periodicity to the fluctuations, although wet or dry habitat conditions influenced by the El Nino-Southern Oscillation may have a role in these fluctuations, as they do in the Galapagos. For example, in the 11+ year population study of the Large Cactus Finch (Geospiza conirostris) by

Grant and Grant (1989), climatic extremes associated with El Nino and the subsequent population extremes tended to recur at intervals of four years or more. Indeed, unusually large amounts of rainfall during 1987 in the Galapagos coincided with a drought year on Laysan Island (Grant and Grant 1989, see Chapter I).

The Laysan Finch population has tended to have peaks and troughs of short duration, and does not apppear to maintain a stable population level. For this species, in a habitat without emigration opportunities, extreme population variability may be the rule rather than the exception

(Dennis et al. in press). During and immediately after years of poor weather (e.g. drought, high winds), density-dependent mortality due to limited resources, coupled with poor reproductive performance, probably causes the population to crash. In years of abundant rainfall, the carrying capacity of the Island increases; abundant resources can promote high survival and high reproductive success, yielding a high population. In some situations, severe weather may also be associated with high density-independent mortality (e.g. severe flooding, or storms that prevent all the finches from foraging; see Chapter I). Rainfall may be useful as a population predictor, particularly if its temporal distribution is known, along with other weather variables, such as peak wind speed.

202 unfortunately, annual censuses have been made during different months, making it somewhat more difficult to evaluate population changes. The stage of the breeding cycle during the censuses has only been recorded during the multiple 1986 to 1988 counts. If Laysan

Finches commence their breeding based on some external stimulus such as rainfall or the post-rainfall flush of vegetation, then the pattern of population peaks and troughs within a single year may be different in different years. Such variability is suggested by comparison of the population estimates for 1968-1969 versus 1984-1985. Net recruitment seems to have occurred between the months of september 1968 and March

1969, whereas heavy net mortality is indicated between November 1984 and

April 1985.

The 1986 to 1988 finch population estimates indicated a decreasing trend (Figure 4-3). Based on other information from breeding studies

(see Chapter I), this trend is not surprising. There were drought conditions in 1987 and reduced fledging success. Limited food resources not only reduced recruitment but almost certainly decreased survivorship of second-year and adult birds. This probably explains the low population estimates for 1988 (Figures 4-5 and 4-6). The population apparently rebounded during late 1988, 1989 and 1990.

Data from only three years is not adequate to predict long-term population trends. The 23-year span of population estimates suggests that the only "trend" for this Laysan population is erratic fluctuation around a level of about 10,000 finches, which may represent the approximate carrying capacity of the habitat. The mean for the 20 years of available data spanning 1968 to 1990 is 11,044 finches, with S.D.=

203 3,999. These calculations are based on a single census for each multiple-census year, so that each of the 20 years is equally represented. Except for 1987, I chose the post-breeding census for multiple-census years, because in most years, the census was made in the late summer or early fall, when the population is generally high because of recruitment.

In multiple-census years, counts for the breeding censuses (in May) were lower than either the pre- or post-breeding censuses for the respective year (Table 4-1, Figure 4-3). The population "decrease" between pre-breeding and breeding censuses may occur largely because female finches remain on the nest during the approximately 16-day incubation period, while males bring food to them. Also, females frequently sit in the empty nest the day before the first egg is laid, and brood the chicks constantly for the first few days after hatching

(see Chapter I). Other studies have shown reduced detectability in birds during certain stages of the breeding cycle (Best and Petersen

1982) • The Laysan counts are probably explained by Di.ehl' s (1981) conclusion that detectability was lowest during incubation. This change

in detectability during the incubation and early brooding stage of the breeding cycle may cause serious underestimates of the true population

size. The post-breeding censuses increased relative to the breeding censuses, not only because females had left their cryptic nests, but also because chicks had fledged.

The Laysan Finch population apparently has a high recruitment potential; the July 1986 post-breeding census data indicated that at

least 23% of the population (based on all birds seen on transects) were

204 recruits. Also, at least two clutches per pair may be produced in some years, and an average of 0.9 fledglings were produced per nest. It is unlikely that this potential is realized every year, as evidenced by the population decline between 1987 and 1988, and the scarcity of second­ year birds in 1988 (Morin pers. obs.). Only a small number of birds, originally banded as fledglings in 1986-1988, have been resighted a year or more later (see Chapter I). This suggests that low survivorship of

HY birds after gaining independence from their parents may occur in some years, or that few HY birds are resighted, or some combination of these two possibilities. I believe that survival of HY birds is often, but not always, low over the winter months (September to February). The

1986-1988 data were collected during an overall population decline, so the poor HY survival from those three years may not be representative.

The overall population climb from August 1988 to October 1990 suggests very strongly that HY survival was much better in some or all of those years. The population estimate from October 1990 is difficult to explain in relation to the July 1989 estimate unless overall survivorship, and probably HY survivorship, was good. However, if food was abundant, higher-than-average fledging success and >2 clutches per pair may have contributed to this high population estimate. comparisons Among Individual Counters

Individual differences among counters in detecting birds have been shown to be a significant problem in other avian population studies

(Kepler and Scott 1981, Verner and Milne 1990). Besides experience with birds in general, these individual differences may be attributable to

205 aural acuity and familiarity with the species (Faanes and Bystrak 1981), carelessness and fatigue (Robbins and stallcup 1981), and visual acuity, training, and motivation (Kepler and Scott 1981). Screening and training prospective counters can reduce but not eliminate this individual variation (Kepler and Scott 1981).

Burnham and Anderson (1984) considered that variable distance line transects can eliminate counter as well as other types of bias, as long as g(O) = 1 (i.e., all birds on the centerline of the transect are detected). They also considered that strip transect counts are often negatively biased when these counter (and other) effects cause some objects in the strip to remain undetected. The detection function for

Laysan Finches will be considered later in this discussion.

Within each census, sample sizes (number of transects) per counter varied from 20 transects per person to 60 transects per person, depending upon the number of counters participating in the 120-transect census. Real differences among counters should be more difficult to detect as the number of transects per counter decreases, but there was no such trend in these data (Table 4-3), perhaps because differences among counters were generally large. There was significant variability among counters for every census except one (May 1987), which was the only census where all of the observers were experienced (Table 4-3).

However, experience does not appear to be the overriding factor influencing the variability among counters. When birds per transect were compared between experienced versus inexperienced counters within each census, no significant differences were found except for the August

1988 census (Table 4-3). However, the August comparison was confounded

206 by the fact that it was the only census with only two counters, and consequently effects due to other individual differences and effects due solely to experience cannot be separated. These results suggest that experience (as narrowly defined in this paper) may sometimes explain some count variability, but that other attributes which vary among individuals (e.g. visual acuity, motivation, carefulness, training, or previous experience in counting other animals) may be as important, or more so.

Comparisons Among Vegetation Types

Another potential source of variation among counts was the effects of the vegetation on the particular transects. Population estimates that involve sampling in more than one habitat type frequently employ a stratified sampling scheme, if the organism has been shown to occur in different densities in the different vegetation types. The average number of finches per transect varied in relation to the transect vegetation only during the four non-breeding censuses (Table 4-2). In the post-breeding censuses, Laysan Finches were densest in the bunchgrass association transects, second densest in the viney association transects, and least dense in "other" transects. In the pre-breeding censuses, finches were densest in the viney association transects, but the vegetation type with the second and third highest density was not consistent between censuses. There was a slight seasonality to finch vegetation preferences, which was probably due to the changing phenology of the different plant species in the bunchgrass and viney habitats throughout the seasons. The finches probably shifted

207 their primary foraging focus to follow the seasonal insect abundance and fruiting and seeding of plants. Stratified sampling in these two major vegetation types may be useful during the non-breeding seasons.

However, the stage of breeding would need to be known during each census

(which it usually is not), and the total area of each vegetation type would need to be known with more accuracy than it currently is.

None of the breeding censuses showed a significant difference in the number of finches per transect relative to the vegetation type, suggesting that finches were rather evenly distributed throughout the vegetation. The lower coefficients of dispersion during the breeding censuses suggest that finches are less clumped during the height of the nesting period when eggs and nestlings are present. This distributional pattern could allow for more optimal resource utilization and less intra-specific interference among breeding birds. Although suitable nest substrates per unit area seemed to be the same in the bunchgrass and the viney associations, nest densities were higher in the viney association (see Chapter II). This is not contradictory to the conclusion that finches seemed more evenly distributed during the height of the nesting period, because more males and sub-adults (but not females on nests) were seen during the breeding transect counts.

Breeding males do not defend a territory around their nests, and usually forage away from their nests, utilizing a variety of different plant associations.

208 Evaluation of Transect Length

Combining nearby transects into groups of four produced normally distributed data for 1986 and 1987, but effectively reduced the original

120 transects to 30 artificial transects of length 366 m (1200 ft).

Hanowski et ale (1990) found that with equal effort, they were able to detect smaller differences in passerine counts using short transects with large sample sizes than when they used long transects and small sample sizes. For this reason, lengthening our transects and decreasing the sample size in the field probably cannot be justified, unless more of these lengthened transects are sampled in future censuses. In low population years (e.g. 1988), it is not clear what combination of transect length and number might be required to obtain normally distributed data. Consequently, after groups of four adjacent transects were combined, I did not attempt to statistically compare the normal data from 1986 and 1987 with the non-normal 1988 data. Ultimately, the value of using normally distributed data rather than non-normal data is not a sufficient reason for changing the current census methodology, especially because increasing the sampled area (e.g. by lengthening the current transects) would significantly increase the human impacts on this fragile ecosystem.

Assumptions and Evaluation of the strip Transect Method

Verner (1985) proposed that the most damaging violations of strip transect assumptions probably occurred if all birds within a strip were not detected, and if birds moved into or out of a strip in response to a moving observer. The assumption that all birds within a strip are detected is an extension of the assumption for all line transects that

209 every bird exactly on the transect centerline is detected. If this assumption fails, double sampling or a priori information should be used to develop a correction factor (Burnham and Anderson 1984).

The population estimates in Table 4-1 suggest that for Laysan

Finches, a correction factor of 7/5 may be appropriate during censuses done during breeding peaks, since the 1986 breeding census of 10,333 birds "corrected" by this factor yields the pre-breeding estimate of about 14,460 birds. There were no bird die-offs to explain such a large and unexpected "drop" in population. The April and May 1987 population estimates suggest a declining population, but this "decline" corresponds to the increasingly cryptic behavior of the female finches as they approach the peak nesting period (Morin pers. obs.), which actually occurred in early- to mid-June in that year (based on counts of simultaneously active nests; see Chapter II).

correcting the May 1988 breeding estimate at the peak of nesting

(confirmed by simultaneously active nests; see Chapter II) with the 7/5 factor predicts a pre-breeding population estimate of 7,281 birds, with at least 2,080 breeding females (the difference between the hypothetical pre-breeding estimate and the breeding estimate). Since 149 fledglings were produced from 165 nests (successful and unsuccessful nests combined) during the 1986-1988 breeding study, each nest yields an average of 0.90, or approximately 1.0, fledgling per nest. (There appears to be little difference among years, since mean number of fledglings per nest was 0.87, 0.93, and 0.93 for 1986, 1987, and 1988, respectively.) If one fledgling is allowed for each breeding female in

1988, then 5,201 (birds detected during the breeding. census) plus 2,080

210 (females on nests which went uncounted during the breeding census) plus

2,080 new recruits (one fledgling per nest) is equal to 9,361 post­ breeding birds, which is extremely close to the actual 1988 post­ breeding estimate (Table 4-1). However, the 7/5 correction factor should be used with extreme caution, since it is only a rough estimate based on two years of data (1986 and 1988). Application of the correction factor also requires that the stage of the breeding cycle be known with certainty during a census; this requirement is difficult to meet without lengthy field studies.

In this study, the low population densities obtained during the May breeding censuses suggest that at least sometimes the first assumption

(that all birds within the strip are detected, or g(x)=l), is not met

(Table 4-1). Indeed, if females are on nests within the bunchgrass

(Eragrostis) clumps, they often will not flush and hence will remain undetected. The bunchgrass frequently occurs in thick clumps up to a meter wide and almost as tall. Even finches simply resting inside a clump will sometimes fail to flush, and may only be detected by audible cues, if at all.

Verner's second and third assumptions - that no bird moves into or out of a strip transect in response to a moving observer, and that no bird is counted more than once - may be restated in more realistic terms: that no bird moves out of the strip without first being detected and counted, that birds moving into the strip during the count, if detected, are not counted, and that the counter does not recount any birds. Regular line transect counts make the same assumption that birds do not move in response to a counter before being detected (Burnham and

211 Anderson 1984). The assumption of birds as relatively stationary objects is frequently violated in practice. Conant et al. (1981) found evidence that Nihoa Finches (Telespiza ultima) were attracted to stationary counters during variable circular plot counts but not to moving counters during transect counts. The finches presumably moved into the circular plots prior to detection. They seemed to be attracted to seabird eggs that remained exposed near the stationary counters for longer periods of time than with moving counters. Nihoa Finches, as well as the closely related and biologically similar Laysan Finches, will prey on eggs of certain seabird species. It is reasonable to believe that Laysan Finches may also be attracted to stationary counters in some areas with nesting seabirds but not especially attracted to moving counters during strip transects.

David (1981) suggested that sampling durations in bird studies should be restricted to perhaps 3 to 5 minutes, mainly in an attempt to reduce the number of recounted birds. If birds are stationary, controlling for count duration becomes unnecessary. The more birds move, the more likely they are to be recounted. This is especially important in mobile, high density populations such as the Laysan

Finches, where a uniform count duration is essential for controlling census effort. There are, conservatively estimated, 53.5 finches per vegetated hectare on Laysan (10,000 finches/ 187 vegetated ha), most of which are not color banded. This high density makes it difficult to keep track of recounted birds unless the transect count duration is short. However, the vast majority of Laysan Finches are detected by sight in the low, open vegetation. Recounting individuals becomes more

212 of a problem in studies using primarily aural detection. The percent of recounted finches is probably as low as most strip transect censuses of high density populations, and is possibly lower than many studies.

Verner's fourth assumption - that birds are assigned as in or out of the strip without error - is hard to evaluate for Laysan.

Fortunately, the width of the strip transect is small, and the vegetation is generally low and open, reducing the probability of incorrectly classifying a bird as in or out of the strip. However, the length and width of the strip are estimated for each transect, and every estimation involves error. Ideally, each strip transect count should receive identical effort; length, width, and transect duration should be uniform. Individual variations in strip width, length, and count duration probably account for a large amount of the variation among counters.

The final assumption, which is identical for all types of line transects, is that each bird detection is an independent event.

However, the CD (Table 4-2) for all censuses in all years was greater than 2, and any number greater than 1 indicates a clumped distribution.

The assumption of independence was least violated during the May breeding censuses and the single April census. Differences in densities among the three habitat classifications were also nonsignificant during the breeding censuses, whereas they were significantly different during pre- and post-breeding counts. These changes in distribution probably reflect differences in social behavior and habitat use between breeding and non-breeding periods. However, Burnham et al. (1980) said that as long as the transect lines are randomly located, it is not necessary to

213 assume that the objects themselves are independently distributed throughout the area.

A review of the literature suggests that total mapping (Verner

1985, DeSante 1981) is the most accurate avian census method, and spot mapping is often assumed to be the second most accurate. Where estimates are not made by total mapping, Verner believes that most estimates for avian species have not been shown to be accurate either absolutely or even as a fraction of the absolute population, since most studies do not standardize the estimate against total mapping. Laysan

Finch censuses have not been standardized by total mapping. Total mapping or even spot mapping would be difficult to carry out on Laysan

Finches, since the percent of banded birds is relatively low and the population is dense. Paired birds defend only the nest site itself; they do not defend a feeding territory, but forage over a large area, where they intermix with many other pairs. So-called "floaters" are very difficult to identify in such a situation. Total mapping is also the most time-consuming and hence the most costly of all estimation techniques.

For the congeneric Nihoa Finch, Conant et ale (1981) compared three variants of the line transect, including strip transects. They concluded that variable distance line transects gave the best estimate: their strip transects yielded a similar population estimate, but with larger 95% confidence limits.

Variants of the line transect method are used widely throughout avian census studies (Brennan and Block 1986, Rotella and Ratti 1986,

Bollinger et ale 1988, Conroyet ale 1988, Hanowski et ale 1990),

214 despite the notion that transect counts of highly mobile animals (e.g. passerines) may seriously violate the assumptions necessary for valid population estimates (Burnham et al. 1980, David 1981, Verner 1985).

The distinction between census and index methods is probably blurred; there is instead a continuum of population estimates based on relatively more or fewer violations of assumptions for each different method.

Fortunately, absolute population estimates are often unnecessary. Emlen

(1981) suggested that indices of relative abundance are often adequate or even preferable (e.g. because of cost/benefit ratio) for detecting population changes. This is the case with Laysan Finches on Laysan.

The strip counts may be viewed as a refined, fairly precise index approximating the Laysan Finch population, which needs a correction factor during the height of the breeding season. Although some assumptions of the method are violated to some degree, most of the violations can be further reduced in magnitude by the careful choice of counters and control of census effort.

MANAGEMENT IMPLICATIONS AND RECOMMENDATIONS

There are over 20 years of Laysan Finch census data for Laysan

Island, all gathered using a strip census technique. Due to the extreme remoteness of the island, the usually restricted available time on site, the limited availability of trained personnel, and the necessity to minimize impacts on the biota, a significantly more time-consuming census method is not feasible. The precision of the estimates is probably adequate for the purpose of monitoring gross population changes, provided that the census is conducted at least yearly. The

215 census activities have an impact on the finches as well as other biota,

and for this reason more frequent censuses should generally be

discouraged. Counters may affect the finches directly by accidentally

collapsing seabird burrows onto finches that have entered the burrows

during their foraging, or by stepping directly on finches or nests

concealed beneath the vegetation underfoot. Also, counters have

inadvertently helped to spread the seeds of introduced pestiferous

plants, such as Cenchrus echinatus, which competes for space and reduces

the regeneration of native plants important to finches, such as the

bunchgrass Eragrostis.

Based on results reported here and input from census participants

over the years, I offer the following general management

recommendations. (1) Finches on Laysan should never be fed by humans,

either accidentally or deliberately, since such experience habituates

finches to approach humans during censuses (i.e. searching for food),

thus inflating transect counts. (2) Researchers should limit their

activities near those seabird species whose eggs are vulnerable to finch

predation (or conduct their activities at night, when finches are not

active), so that finches will not learn to follow humans in those areas;

such following also could inflate counts. (3) Aerial photographs of

Laysan should be made at least every five years, and the total vegetated

area quantified to see whether it varies substantially from 187 ha. The

actual area of each vegetation type should be quantified, so that

stratified sampling could be done, if desired or required.

I offer the following specific recommendations for census

methodology: (1) Prior to any census, the stage of the breeding cycle

216

------_..- .. should be ascertained by nest searches and behavioral observations, as described elsewhere (see Chapter I). Although the finches had breeding peaks in May from 1986 to 1988, they may have more than one peak of breeding in some years. Censuses done during the height of the breeding season satisfy the assumption of independence better than censuses during the non-breeding season because breeding finches are most evenly dispersed throughout the vegetation and are the least clumped (Table 4­

2). A correction factor of 7/5 is suggested for censuses done during peak breeding periods, since many females on nests are not counted.

However, censuses done during the post-breeding season better satisfy the assumption that all birds within the strip are detected, even though they are more clumped (possibly due to family groups and fledgling aggregations).

(2) Use the same counters whenever possible; this should help to reduce count variability (Table 4-3). If inexperienced persons must be used, they should do several practice transects the day before the actual census so that any questions or concerns about methodology may be resolved. However, my data suggested that previous experience alone is not an important source of count variability.

(3) Have each counter carry a lightweight pole, with 2.5 m (one­ half the transect width) marked on it so that strip width can be accurately estimated. In spite of the awkwardness of walking with a long pole, I received favorable comments from counters who did their transects with poles.

217 (4) Prior to the actual census, each counter should practice walking a measured course and determine the average number of paces he or she uses for a 91.4 m transect.

(5) Vocalizations should be counted if the sound can definitely be assigned to the inside of the strip and the finch is not otherwise seen.

Aural detections have rarely been recorded during counts, probably because of difficulty in assigning location rather than scarcity of vocalizations. Also, if a bird is heard, it is almost always seen.

(6) Counters should spend three minutes per transect. Practicing with timers before the count and using timers during the count may help to standardize this. This pace is approximately equal to the "slow, easy walk" which pre-1986 counters described.

(7) Counters should count birds up to 2.5 m to the side and in front of themselves (although not beyond the end of the transect), but not behind themselves. If a bird is seen leaving this 2.5 m radius, it should be counted, but if it flies into this 2.5 m radius from outside, it should not be counted. No bird should be recounted, and birds that fly over the transect should not be counted. Inexperienced counters must be briefed about this protocol with special care.

The usual restrictions on time and weather (i.e. count in the morning, and do not count during heavy rain or wind) should also be followed.

218 LITERATURE CITED

Best, L. B. and K. L. Petersen. 1982. Effects of stage of the breeding

cycle on Sage Sparrow detectability. Auk 99: 788-791.

Bollinger, E. K., T. A. Gavin, and D. C. McIntyre. 1988. Comparison of

transects and circular-plots for estimating Bobolink densities. J.

Wi1dl. Manage. 52: 777-786.

Brennan, L. A. and W. M. Block. 1986. Line transect estimates of

Mountain Quail density. J. Wildl. Manage. 50: 373-377.

Brock, V. E. 1951. Laysan Island bird census. 'Elepaio 12: 17-18.

Bull, E. L. 1981. Indirect estimates of abundance in birds. Pp. 76­

80. In Estimating numbers of terrestrial birds. C. J. Ralph and J.

M. Scott (eds.). Stud. Avian Biol. 6.

Burnham, K. P. and D. R. Anderson. 1984. The need for distance data in

transect counts. J. Wildl. Manage. 48: 1248-1254.

Burnham, K. P., D. R. Anderson, and J. L. Laake. 1980. Estimation of

density from line transect sampling of biological populations.

Wildl. Monographs No. 72. 202 pp.

Conant, S., M. S. Collins, and C. J. Ralph. 1981. Effects of observers

using different methods upon the total population estimates of two

resident island birds. Pp. 377-381. In Estimating numbers of

terrestrial birds. C. J. Ralph and J. M. Scott (eds.). Stud. Avian

BioI. 6.

Conroy, M. J., J. R. Goldsberry, J. E. Hines, and D. B. Stotts. 1988.

Evaluation of aerial transect surveys for wintering ~_merican black

ducks. J. Wildl. Manage. 52: 694-703.

219

------David, F. N. 1981. Statistics for the birds. Pp. 566-569 In

Estimating numbers of terrestrial birds. C. J. Ralph and J. M. Scott

(eds.). Stud. Avian Biol. 6.

Dawson, D. G. 1981. Counting birds for a relative measure (index) of

density. Pp. 12-16. In Estimating numbers of terrestrial birds.

C. J. Ralph and J. M. Scott (eds.). stud. Avian Biol. 6.

Dennis, B., P. L. Munholland, and J. M. Scott. In press. Estimation of

growth and extinction parameters for endangered species. Ecological

Monographs.

DeSante, D. F. 1981. A field test of the variable circular-plot

censusing technique in a California coastal scrub breeding bird

community. Pp. 177-185. In Estimating numbers of terrestrial

birds. C. J. Ralph and J. M. Scott (eds.). Stud. Avian Biol. 6.

Diehl, B. 1981. Bird populations consist of individuals differing in

many respects. Pp. 225-229. In Estimating numbers of terrestrial

birds. C. J. Ralph and J. M. Scott (eds.). Stud. Avian Biol. 6.

Emlen, J. T. 1981. Summary of the symposium. Pp. 575-576. In

Estimating numbers of terrestrial birds. C. J. Ralph and J. M. Scott

(eds.). Stud. Avian Biol. 6.

Faanes, C. A. and D. Bystrak. 1981. The role of observer bias in the

North American breeding bird survey. Pp. 353-359. In Estimating

numbers of terrestrial birds. C. J. Ralph and J. M. Scott (eds.).

Stud. Avian Biol. 6.

Ely, C. A. and R. B. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll Res. Bull. 171. 361 pp.

220

------_.- Grant, B. R. and P. R. Grant. 1989. Evolutionary dynamics of a natural

population: the Large Cactus Finch of the Galapagos. Univ. of

Chicago Press, Chicago. 350 pp.

Hanowski, J. M., G. J. Niemi, and J. G. Blake. 1990. Statistical

perspectives and experimental design when counting birds on line

transects. Condor 92: 326-335.

Hussel, D. J. T. 1981. The use of migration counts for monitoring bird

population levels. Pp. 92-102. In Estimating numbers of

terrestrial birds. C. J. Ralph and J. M. Scott (eds.). Stud. Avian

BioI. 6.

Kepler, C. B. and J. M. Scott. 1981. Reducing bird count variability

by training observers. Pp. 366-371. In Estimating numbers of

terrestrial birds. C. J. Ralph and J. M. Scott (eds.). Stud. Avian

BioI. 6.

Lamoureux, C. H. 1963. The flora and vegetation of Laysan Island.

Atoll Res. Bull. 97: 1-14.

Morin, M. P. In prep. Laysan Finch nest characteristics, nest spacing,

and reproductive success in two vegetation types.

Morin, M. P. In prep. The breeding biology of an endangered Hawaiian

honeycreeper, the Laysan Finch (Telespiza cantans).

Newman, A. L. 1988. Mapping and monitoring vegetation change on Laysan

Island. M. A. thesis, Geography Dept., Univ. of Hawaii, Honolulu.

234 pp.

Nichols, J. D., B. R. Noon, S. L. Stokes, and J. E. Hines. 1981.

Remarks on the use of mark-recapture methodology in estimating avian

221 population size. Pp. 121-136. 1n Estimating numbers of terrestrial

birds. C. J. Ralph and J. M. Scott (eds.). stud. Avian Biol. 6.

Olson, S. L. and H. F. James. 1982. Prodromus of the fossil avifauna

of the Hawaiian Islands. Smithsonian Contrib. Zool. 365. 59 pp.

Pollock, K. H., J. D. Nichols, C. Brownie, and J. E. Hines. 1990.

Statistical inference for capture-recapture experiments. Wildl.

Monographs No. 107. 97 pp.

Reynolds, R. T., J. M. Scott, and R. A. Nussbaum. 1980. A variable

circular-plot method for estimating bird numbers. Condor 82: 309­

313.

Robbins, C. S. 1981. Bird activity levels related to weather. Pp.

301-310. In Estimating numbers of terrestrial birds. C. J. Ralph

and J. M. Scott (eds.). Stud. Avian Biol. 6.

Robbins, C. S. and R. W. Stallcup. 1981. Problems in separating

species with similar habits and vocalizations. Pp. 360-365. In

Estimating numbers of terrestrial birds. C. J. Ralph and J. M. Scott

(eds.). Stud. Avian Biol. 6.

Rotella, J. J. and J. T. Ratti. 1986. Test of a critical density index

assumption: a case study with gray partridge. J. Wi1dl. Manage. 50:

532-539.

SAS Institute Inc. 1988. SAS/STAT User's Guide, Release 6.03 Edition.

SAS Institute Inc., Cary, NC.

Scott, J. M. and F. L. Ramsey. 1981. Length of count period as a

possible source of bias in estimating bird densities. Pp. 409-413.

In Estimating numbers of terrestrial birds. C. J. Ralph and J. M.

Scott (eds.). Stud. Avian Biol. 6.

222

------_. -_. -----~------" Sincock, J. L. and E. Kridler. 1977. The extinct and endangered

endemic birds of the Northwestern Hawaiian Islands. Unpubl. Ms., U.

S. Fish and Wildlife Service, Honolulu. 111 pp.

Skirvin, A. A. 1981. Effect of time of day and time of season on the

number of observations and density estimates of breeding birds. Pp.

271-274. In Estimating numbers of terrestrial birds. C. J. Ralph

and J. M. Scott (ede,), Stud. Avian Bio1. 6.

Sokal, R. R. and F.J. Rohlf. 1981. Biometry. W. H. Freeman and Co.,

New York. 859 pp.

U. S. Fish and Wildlife Service. 1984. Recovery plan for the

Northwestern Hawaiian Islands passerines. U. S. Fish and Wildlife

Service, Portland. 66 pp.

Verner, J. 1985. Assessment of counting techniques. Pp. 247-302. In

Current ornithology. Vol. 2. R. F. Johnston (ed.). Plenum Press,

New York.

Verner, J. and K. A. Milne. 1990. Analyst and observer variability in

density estimates from spot mapping. Condor 92: 313-325.

Verner, J. and L. V. Ritter. 1988. A comparison of transects and spot

mapping in oak-pine woodlands of California. condor 90: 401-419.

223 TABLE 4-1. Census summary for 1986 to 1988.

STAGE OF CENSUS POPULATION 95% MEAN COUNT STANDARD BREEDING DATE ESTIMATE CONFIDENCE PER DEVIATION CYCLE INTERVAL TRANSECT (RANGE) PRE-BREED 27 FEB 86 14,468 ±2,257 3.55 3.02(0-14)

BREEDING 28 MAY 86 10,333 ±1,796 2.53 2.44(0-10)

POST-BREED 23 JUL 86 14,786 ±2,360 3.63 3.20(0-18)

PRE-BREED 15 APR 87 11,659 ±1,812 2.86 2.46(0-13)

BREEDING 28 MAY 87 10,775 ±1,783 2.64 2.42(0-10)

BREEDING 28 MAY 88 5,201 ±1,211 1.28 1.64(0-8)

POST-BREED 15/16AUG 88 9,349 ±1,915 2.29 2.60(0-14)

224

------TABLE 4-2. Comparisons of birds per transect for the three vegetation classifications during each census, and coefficients of dispersion.

CENSUS MEAN COUNT CHI SQUARE1 DEGREES P VALUE COEFFICIENT DATE PER OF OF TRANSECT FREEDOM DISPERSION 27 FEB 86 3.55 14.332 2 0.0008 2.57

28 MAY 86 2.53 2.8088 2 0.2455 2.35

23 JUL 86 3.63 6.3659 2 0.0415 2.82

15 APR 87 2.86 10.752 2 0.0046 2.12

28 MAY 87 2.64 2.6351 2 0.2678 2.22

28 MAY 88 1.28 1.0936 2 0.5788 2.10

15/16 AUG 88 2.29 15.861 2 0.0004 2.95

1 Kruskal-Wallis test to determine differences in mean transect counts for the three different vegetation classifications.

225 TABLE 4-3. Within each census, comparisons of numbers of birds per transect by individual counters, and by experienced vs. inexperienced counters.

CENSUS NUMBER OF TEST1 DEGREES P VALUE TEST2 P VALUE DATE COUNTERS STATISTIC OF STATISTIC 2) (INEXP) (X FREEDOM (Z) 27 FEB 86 4(3) 13.068 3 0.0045 0.8174 0.4137

28 MAY 86 6(2) 15.336 5 0.0092 -1. 6105 0.1073

23 JUL 86 5(3) 12.170 4 0.0161 -0.9883 0.3230

15 APR 87 6(3) 13.738 5 0.0174 -1. 6170 0.1059 3 28 MAY 87 6(0) 10.144 5 0.0713 NO TEST

28 MAY 88 5(2) 10.285 4 0.0359 -1. 5925 0.1113

15/16 AUG 88 2(1) Z=-4.11l0 0.0001 -2.3208 0.0220

1 Kruskal-Wallis or Mann-Whitney(Aug 88 only) tests for significant differences in mean counts among observers within each census.

2 Mann-Whitney test for significant differences in mean counts between experienced and inexperienced observers within each census.

3 No test - All observers experienced.

226 Figure 4-1

Vegetation map of Laysan Island.

1 km I

II Scaevola Shrubland ~ Eragrostis Association

[] lpomoea-Boerhavia-Sicyos Association

~ Pluchea Association mm Sesuvium-Heliotropium Association 227 Figure 4-2 Grid map used to locate transects on Laysan Island.

N

Transect Site

Water or Bare Lake Bed

o 1 km I 228 PR = PRE-BREEDING 18 BR = BREEDING 17 PO = POST-BREEDING 16 (/) 15 Cl z 14 -c (/):::» 13 ~ 12 I- 11 PR PO z 10 z o 9 PR ~ ...J 8 BR BR BR ~ 7 ~ 6 PO 5 4 3 ...... ------FEB MAY JUL APR MAY MAY AUG 86 86 86 87 87 88 88

DATE OF CENSUS FIGURE 4-3 Laysan Finch population estimates and 95% confidence intervals.

229 220 - ~ EXPECTED BY 200 - POISSON

180 - c=J OBSERVED

160 - N = 837 TRANSECTS

140 - >­o z 120­ LLJ ::> CJ LLJ 100- 0::: - u, 80 - - 60 - .....

40 -

20 - - _..... 0-- - -- _..... _-.I Jl n o 2 3 4 5 6 7 8 9 10+ FINCHES PER TRANSECT

FIGURE 4-4 Distribution of the number of finches counted per transect for all seven censuses combined.

230

--- --~- ~.. --._--_._ .._- . __ ... __ . 500 • • en 450 I- • Z ::> 0 .~ u 400 • 0 A w :l£ -cZ -, a::: B I.J.. 350 0 z -c w ::!: 300 ~ 250 co co CD eo eo IX) IX) '"IX) '"eo eo eo eo .....J m ~ >- C) >- :::> w 0- -c ~ ::> -c ..., IJ.. -c ::!: ::!: -c ::!: DATE OF COUNT

FIGURE 4-5

Tukey grouping using ranks for count data. Means with the same letter are not significantly different. The number of transects per date was 120 except for FEB 86 which was 117.

231 25 en 0 z -c en 20 :::J 0 ::I: I- z 15 en w I--c ::E I-en 10 w z 0 I- ~ 5 :::J n, 0 n, o ..... ~.,,/,r.I------IDmO..-NI"l'OtIOCO,...IOO) 1''l~1I110101O,...,...lOlOo)O 1010,...,...,...,...,...,...,...,...,...,... IOIXl 10101010 10 ID 1010 100) Q..O:::CJQ..Q....J..Je"e"e"e,,>­ ..J>O:::ID>-..Jo:::>->-e"..J1­ 1AJ<::JI1JIAJ::J::J::J::J::J::J< ::JOQ..IAJ<::JQ.«::J=>O m::::E

FIGURE 4-6

The population estimates ±95% confidence intervals of Laysan Finches on Layson Island. The 1986-1988 estimates are extensively analyzed in the text. Intervals on the date axis are not regular.

232 GENERAL CONCLUSIONS AND COMMENTS

The breeding ecology of the Laysan Finch (Telespiza cantans) was

studied during three field seasons on Laysan Island in the Hawaiian

Archipelago. This endangered species had been little studied because of

its remote location. However, it is of considerable interest because it

is thought by some (Sibley 1970, Raikow 1977, van Riper 1978) closely to

resemble the ancestral type from which other Hawaiian Honeycreepers

arose. It is closely related to other, much rarer or extinct members of

the finch-billed Honeycreeper tribe (Psittirostrini), making it an ideal

surrogate study subject. The population of Laysan Finches on Laysan

Island also provides a good system for general ecological, life history,

and population dynamics studies because it is a closed population in a

simple ecosystem limited to an area of manageable size. In addition to

studying the breeding ecology and population size of this species, my

research considered several hypotheses regarding hatching asynchrony

that have been proposed by workers in other systems in relation to

Laysan Finch reproductive success and nestling growth.

As a result of past habitat degradation on Laysan, two other

endemic passerine species and an endemic rail are now extinct. The

interspecific influences (e.g. competition, and in the case of the rail,

possibly predation) that these birds may previously have exerted no

longer occur, eliminating some of the factors that probably limited the

population in the past.

Several aspects of the Laysan Finch's breeding biology, such as its

ability to produce multiple broods within a single breeding season,

233

------.. demonstrate why this species rebounded so quickly from the severe population bottleneck that occurred in the early part of this century

(Ely and Clapp 1973). Limiting factors that currently regulate this closed population are not fully understood, although this research presented evidence that two major limiting factors are weather (both direct and indirect effects) and intraspecific predation.

BREEDING BIOLOGY AND POPULATION LIMITING FACTORS

The data gathered during the three field seasons characterized this

Hawaiian Honeycreeper as a monogamous species, with at least some year­ to-year mate fidelity. A small Type Two territory (mating and nesting only: Welty 1982) was maintained, and both sexes appeared to guard their mates. A pair often built nests in close proximity for renest attempts within a year or in successive years. If the first nest in a breeding season was successful, at least one more clutch was possible under favorable environmental conditions. Within a season, at least two renests were possible by the same pair if the initial nest failed.

Fledgling development and fledgling dependency times were prolonged relative to many passerines of similar size, suggesting that there is an advantage to slow chick growth. Perhaps it allows parents more time to gather food when resources are scarce or far away, as is the case with many species of seabirds.

There were few nests where all eggs hatched. A large percent of eggs disappeared, probably because of intraspecific predation; this may be an important density-dependent limiting factor. Because the female usually began incubation with the first egg, eggs usually hatched

234 asynchronously in the order laid. It is possible that this immediate incubation may, at least in part, be an anti-predation adaptation, where the usual predator is a conspecific. However, predation at the nest by several migratory shorebirds may also occur.

Known-age chicks that fledged differed significantly from those that did not fledge in all measurements taken at age 9/10 days old except for sternum length. Sternum length may provide the best criterion for aging chicks, since it seems to be the body measurement least affected by inadequate nutrition. In nests where three or more eggs hatched, chick weight, and sometimes beak length, varied significantly among different chicks in the hatching order at days 9 and

10 (combined), but five other measurements did not. In nests where only two eggs hatched, regardless of the initial clutch size, there were no differences between first- and second-hatched chicks for any of the seven body measurements at day 9 and 10. In addition, in nests where only two eggs hatched, fledging success was approximately equal for the two chicks; in nests where three or more eggs hatched, the later-hatched chicks had a lower fledging success than the earlier-hatched chicks. At banding, chicks that hatched earlier in the hatch sequence had significantly more furcular fat than later-hatched chicks, suggesting that earlier-hatched chicks were generally in better condition.

Inexplicably, four-egg clutches had relatively more infertile or addled eggs than any other clutch size. The results of this research did not clearly favor anyone of the current hypotheses concerning hatching asynchrony in birds. Instead, several hypotheses (most obviously the

235 brood reduction hypothesis, but also the nest failure and insurance egg hypotheses) could all partially explain the data.

Almost all nests occurred in the two most common vegetation associations (the predominantly bunchgrass, or Eragrostis variabilis association, and the viney, or Ipomoea-Boerhavia-Sicyos association), and almost every nest was in a single plant species, the bunchgrass. No difference in average clutch size per nest or average fledglings produced per nest could be detected between the two vegetation associations during any year. The most common vegetation association

(bunchgrass) did not appear to be the preferred association, based on two years of data on nest densities and nearest-neighbor nest distances.

The density of nests in the viney association was at least twice that of the bunchgrass association, although measured differences in the nest substrates and other nest variables were small when the two associations were compared. The preferred viney association occurs farther inland than the bunchgrass association and is slightly more buffered from storms; during one breeding cycle (early 1986) when a severe storm occurred, this association had the only nests where any nestlings survived to fledge. This kind of random, density-independent mortality is probably a fairly common occurrence in this population. It is hypothesized that the topography of the island has allowed the more protected microhabitat in the viney area to become preferred as a nesting area, although more abundant and varied food resources in that association during some years may also explain its heavy usage.

The current invasion of a non-native bush (Pluchea indica) throughout the preferred nesting area is of concern and should be the

236 subject of further research and control. If this bush continues to

spread around the lake at the same rate that it has over the past 20

years, it will eventually competitively displace most or all of the

productive viney association. Some roosting, but very little foraging

by finches has been observed in these non-native bushes, and only a

single finch nest in the skeleton of a dead bush overgrown with other

native plants has ever been found.

Only 3.5% of 284 fledglings banded in the nest were seen a year or

more later. During a single post-breeding census in the wettest year,

an estimated 22.5% of all birds seen were recently fledged (hatch-year

or HY) birds. Taken together, these data suggest that Laysan Finches

have very low post-fledging survival at least some of the time, and that

the hatching year mortality occurs primarily during the winter months.

Weather - most notably precipitation, but also wind speed and

temperature - fluctuated considerably and unpredictably both within and

among years. Based on vegetation observations, as well as precipitation

measurements, 1986 was classified as a wet year, 1987 as a drought year,

and 1988 as a more representative year. In 1986, a severe storm during

the early peak of breeding caused almost total egg and chick mortality.

In the drought year, the onset of breeding season occurred more than a

month later than in the other two years.

Furcular fat status from a subset of female parents suggested that

females were in poorer condition during the drought year. Fatter

females from all years tended to lay larger clutches and fledge more

chicks per nest. However, there was no relationship of female weight

with mean egg weight per clutch, nor with total egg weight per clutch.

237

------_.. This suggests that female condition but not female size was the

important factor, consistent with Murphy's (1986) results for another

passerine. Mean egg weight per clutch during the drought was

significantly less than in 1986 and 1988, more malformed eggs were laid,

and clutch size tended to be smaller. The reduction in mean egg weight per clutch during the drought was mostly due to lighter eggs in four-egg clutc~es relative to three-egg clutches; in the wet year the opposite

relationship occurred. For all years combined, average-size eggs had the highest probability of producing fledglings, suggesting that there

is stabilizing selection for an optimal egg weight.

Overall, four-egg clutches were most productive, but this was not true in every year. In the drought year, eggs from four-egg clutches produced fledglings at about the same rate as eggs from three-egg

clutches, although four-egg clutches were more productive during 1986 and 1988. The year-to-year variability in fledging success for five-egg

clutches was even more exaggerated. This appears to demonstrate that occasional selective pressure can promote stabilization at a modal

clutch size smaller than the most productive size ever experienced.

This is true because the most productive clutch size varies from year to

year. In years of nutritional stress, if an extra egg (i.e. the fourth

or fifth egg) is produced at the expense ..of all the other eggs (e.g., so that all eggs are reduced in size), then under poor conditions, as many or more offspring will be fledged by those females which lay a smaller

clutch (e.g. three eggs).

These data suggest that serious droughts or other severe weather

events (such as storms) that limit parental food-gathering must occur

238 fairly frequently on Laysan Island. Individual single-year studies of breeding biology can yield very different conclusions about optimal clutch size. This observation led Boyce and Perrins (1987) to emphasize that long-term studies are needed to adequately describe breeding biology, so that mechanisms such as the "bad-years effect" can be detected. Boyce and Perrins demonstrated that in the long term, it is generally advantageous for birds to lay clutches smaller than the most productive size, because birds that lay large clutches in "bad years" have a reduced mean fitness and increased variance relative to birds that lay smaller clutches. Laysan Finch clutch sizes appear to fit this model.

POPULATION ESTIMATES AND EFFECTIVE POPULAT:ION SIZE

population censuses indicated that the population declined from

1986 to 1988, with the sharpest drop between 1987 (the drought year) and

1988. Reduced recruitment due to poor breeding success in 1987, and

increased subadult and adult mortality, mediated by reduced food

resources, are likely explanations. When the peak of the 1987 breeding

season is compared to the 1988 peak, nest densities within the study

area tended to be lower in 1988, and nearest-neighbor nest distances

tended to be higher, although these differences were not significant.

The projected total number of nests on the island during the peak of

breeding activity was more than 450 nests larger in 1987. This strongly

suggests that the population had in fact declined. Taken together, the

data indicate that this population is strongly regulated by the effects

of weather. In this species, with no opportunities for emigration or

239 immigration and only a small amount of available habitat, high population variability may be the rule (Dennis et ale in press).

Analysis of census data suggests that strip censuses made during the height of the breeding season underestimate the true population, because incubating females seldom leave the nest and are therefore not seen by observers. This type of problem has been recognized in other passerine studies (Best and Petersen 1982). A crude correction factor of 7/5 is suggested for estimates made during that stage of the Laysan

Finch breeding cycle. In spite of the drawbacks, strip censuses appear to provide an adequate method for monitoring changes in this population, especially if efforts are made to control for counter variability and counter effort, and if stage of the breeding cycle is accurately determined from other observations.

The effective population size (Ne) is generally one-third to one­ fourth the actual population size. For short-term maintenance, the total Laysan Finch population should be at least 150 to 200 birds, and for long-term maintenance, at least 1,500 to 2,000 birds (Soule 1980).

The lowest total population estimate (May 1988) of about 5,000 finches yields an estimated Ne of 1,250 to 1,666. However, if the 7/5 correction factor (see Chapter IV) is applicable to that May breeding estimate, then the true total population was closer to 7,000 finches, with a minimum estimated Ne of 1,750. Effective population size is also influenced by an unbalanced sex ratio, population fluctuations, and variance in individual reproductive success; all of these occur in this

Laysan Finch population and further reduce the Ne• Although the short-

240 term population maintenance of this bird seems secure, long-term population maintenance is uncertain.

PHYLOGENETIC COMMENTS

Although the results of recent genetic studies are unclear as to the phylogenetic placement of the Laysan Finch in relation to the other

Hawaiian honeycreepers (R. C. Fleischer pers. corom.), starch gel electrophoretic studies have led some to suggest that the Laysan Finch may be congeneric with the Palila (Loxioides bailleui), another finch­ billed Hawaiian honeycreeper (Johnson et ale 1989). The natural history of the. Laysan Finch is similar in many ways to that of the Palila. Both species have extended breeding seasons, tied at least somewhat to food resources, courtship and nestling feeding by regurgitation, mate and nest guarding rather than traditional territory defense, year-to-year fidelity in the nest area, similar age at fledging, an extended period of fledgling dependency, and the ability to raise two broods per breeding season (van Riper 1978, Pletschet and Kelly 1990). However, they differ in modal clutch size, degree of nest sanitation, and participation by the sexes in nest construction.

Eddinger's (1970) breeding biology study of four Hawaiian honeycreepers on Kauai provided possible behavioral evidence about the most likely ancestral group for the Hawaiian honeycreepers. He concluded that two groups (the thraupines and the coerebines) which are now in the Family Emberizidae were the most likely ancestral stock.

This conclusion agrees with the recent suggestion that the Hawaiian honeycreepers are more closely allied with the emberizids (Johnson et

241 ale 1989). Both studies concluded that the ancestral form of the

Drepanidinae had a generalized bill, tongue, and diet.

In contrast, Raikow's (1977) morphological studies and van Riper's

(1978) breeding biology studies concluded that the Hawaiian honeycreepers were most closely related to the cardueline finches.

Sibley's work (1970) with egg-white proteins also supports a cardueline ancestor for the Hawaiian honeycreepers. Sibley even suggested that the ancestral stock of the Hawaiian Honeycreepers could have been closest to the Crossbills (Loxia appv ). Results of subsequent genetic studies by

Sibley and Ahlquist (1982) were in agreement with van Riper and Raikow in suggesting a cardueline ancestor. Thus, conflicting evidence suggests that Laysan Finches may be one of the most derived lineages in the Hawaiian honeycreeper group (Johnson et ale 1989) or may be the lineage most like the primitive ancestor (Raikow 1977). However, recent genetic studies indicate that the Laysan Finch may be neither the most derived ~1.or the most primitive (R. Fleischer, pers. commv ),

The results of my research confirm that the Laysan Finch has many breeding behaviors which closely resemble carduelines. Both carduelines and Laysan Finches: feed their young at long intervals via regurgitation, nest in small territories, obtain their food outside the territory, exhibit courtship and incubation feeding, and maintain nest sanitation until almost fledging time. For both, the female chooses the nest site and builds the nest alone, and only the female incubates, mainly fed by the maLe s- In at least two cardueline species (Redpolls and Crossbills), incubation begins with the first egg and the eggs hatch asynchronously (Newton 1973). The incubation length (13-16 days) and

242 the nestling period (16-25 days) of the Crossbill are very similar to the Laysan Finch. Both the typical cardueline and the Laysan Finch follow the general behavioral sequence: 1) pair formation, 2) nest site selection, and 3) "territory" establishment. For the Laysan Finch, the territory is very small and basically encompasses only the nest or the vegetation immediately surrounding the nest site proper.

Many carduelines alter the onset of nesting when food is plentiful. The Crossbills will even nest in any month, regardless of the temperature, if food is plentiful; food has apparently assumed the major proximate control over breeding (Newton 1973). Food availability has some proximate influence on breeding onset in Laysan Finches, as evidenced by the delayed breeding season during the drought year in this study. However, the degree to which food acts as a proximate control is unclear.

In summary, Laysan Finches appear to have the major behavioral attributes typical of cardueline finches. However, behavioral and life history attributes are generally not the best evidence to use in avian phylogenetic classification. Some other types of evidence about phylogenetic relationships remain conflicting. In order to clarify the relationships, future genetic studies on Laysan Finches should use several species of cardueline finches and emberizids as outgroups.

CONSERVATION

The scientific component of conservation biology is research that describes the dynamics of small populations, their ecological and social interrelationships, and tests hypotheses concerning them. With creative

243 application, these scientific results can be transformed into management and public policy that preserves biodiversity and restores or stabilizes damaged ecosystems.

Islands in general are vulnerable to habitat damage and biodiversity loss because of human activities; the smaller and more pristine the island, the more immediately apparent are such negative impacts. Laysan and the other refuge islands in the leeward chain are relatively pristine when compared to the permanently inhabited main

Hawaiian Islands. The Laysan Finch is more fortunate than many endangered species because it inhabits a remote, protected ecosystem where the opportunities to stabilize and restore damaged native habitat are good. Conservation of species assemblages in situ, where feasible, is by far the preferred strategy (Scott et ale 1988), for evolutionary reasons as well as management practicality.

Like most conservation problems, the maintenance of the finch and its ecosystem is mostly a matter of managing people. Although access to

Laysan is theoretically restricted to researchers and managers, even these few visitors have significant direct and indirect impacts.

Humans, their equipment, and structures have caused direct mortality. Finches that have entered tents, gone under tarps, or into uncovered pit toilets have died of hyperthermia and starvation when unable to find a way out. Finches have been attracted to objects such as uncovered ice-chests and buckets full of fresh water, where they have drowned (Morin 1987). This vulnerability to drowning in fresh water has also been observed in the closely related Nihoa Finch (Amerson 1971).

Infrequently, finches have been accidentally stepped on.

244 Humans also have an indirect negative impact on Laysan Finches through native ecosystem alteration. As on other islands, a serious and

insidious threat to the native ecosystem is the accidental or deliberate

introduction of non-native plants and animals (Loope et ale 1988). Such

introductions sometimes have disastrous consequences, including extinctions (Atkinson 1985, Savidge 1987). In the past, many species have been brought to Laysan by man (Ely and Clapp 1973, Sincock and

Kridler 1977), and several of them pose a threat to Laysan Fin'?_ll~~__ and their ecosystem (e.g. the grass Cenchrus echinatus, the bush Pluchea

indica, the ant Monomorium pharaonis). The introduced plant species displace native plants important to the finches (Morin and Conant 1990;

see Chapter II). The ants have been implicated in the deaths of several

finch chicks, all of which occurred in the vicinity of the tent camp.

The ants are also known to carry disease organisms (story 1986), and are potential disease vectors for all the bird species. The accidental

introduction of rats or other predatory mammals would lead to certain catastrophe for ground nesting bird species, including Laysan Finches.

In the past few years, the U. S. Fish and Wildlife Service has begun investigations into control of pestiferous non-native species on the remote refuge islands, including Laysan. Management of the finch

for the maintenance of a long-term viable population must ensure that human impact on the ecosystem is minimized. Further habitat damage

should be prevented, and current damage should be rectified, by means

such as prioritized removal of aggressive non-native species. Human activities on these fragile islands must also be regulated; pestiferous

species already present are spread within an island and between islands

245 on clothing and equipment. The strict prevention program, which includes freezing, spraying, and washing clothes, equipment and food, should be continued for all visitors to the remote refuge islands.

These precautions definitely reduce the chances of accidentally introducing non-native species. However, the rapid increase in the fishing industry around Laysan and the other leeward islands in recent years makes the probability of unauthorized and accidental landings on the islands very high, greatly increasing the risk of introducing new species, and exposing the Laysan Finch and its ecosystem to a higher likelihood of extinction.

FUTURE RESEARCH

Four areas of potential research emerged from my work. I would recommend prioritizing ecosystem research that might help stabilize and restore the habitat, since results of the other studies will be affected by a degrading ecosystem.

ECOSYSTEM - 1) Development of control and removal methodology for non­ native species, and restoration of native species lost during the 1923 devegetation, if possible.

2) Investigation of the effect produced by foraging Laysan

Finches on vegetation structure and composition, especially in years of high or low population. There are indications that some plant populations (e.g. Scaevola and Tribulus) may be at least partly regulated by finch eating habits.

246 POPULATION- 3) Many, if not most, Laysan Finches in the small, introduced population at Pearl and Hermes Reef are banded; their demography is currently being studied (Conant pers. corom.). In contrast, the large (mostly unbanded) population on Laysan is more difficult to study; it is hard to relocate banded individuals to verify fecundity, mortality, etc. It is probable that selection pressures at the two atolls are different, and hence population processes may diverge. An intense, long-term banding program at Laysan would probably reveal striking contrasts in life history parameters (e.g. survival between years, and age of first breeding in males) in comparison to data from Pearl and Hermes.

4) Monitoring population fluctuations is an important objective for the managers of an endangered species. On the other hand,

Laysan Finch censuses on Laysan have negative impacts on the ecosystem, so these two considerations must be carefully weighed. My data showed that the population estimates seemed to decrease during the height of the nesting season, and I suggested a rough correction factor for census estimates made during that breeding stage. This correction factor could be verified and refined by additional data from multiple-census years.

Because onset of breeding is variable among years, such a project would require that breeding stage be continuously and closely monitored, as I did during my field work. Other population data, such as seasonal and yearly recruitment, and general age-class structure, could be gathered during censuses.

247 BEHAVIOR 5) Dominance behavior within this species is poorly understood, but the presence of delayed plumage maturation and variable bill size between and within the sexes suggest that many interesting observations remain to be made.

6) Examination of intra- and interspecific predation on finch eggs in the nest as a population limiting factor. The relative importance of these is unknown. Research on the Laysan Finches at Pearl and Hermes Reef might initially be most fruitful, since nests there are more exposed (Morin and Conant 1990).

PHYSIOLOGICAL AND GENETIC-7) Clarify the finch's relationship to other

Hawaiian honeycreepers and to outgroups such as the carduelines and emberizids with more detailed genetic studies.

8) Investigate physiological and behavioral adaptations to environmental extremes. For example, how do ambient temperature, nest temperature, and the number of chicks per clutch interrelate, and what are the SUbsequent survival rates under various regimes? This may provide insight into natural selection on the upper limits for clutch size.

248 LITERATURE CITED

Amerson, Jr., A. B. 1971. The natural history of French Frigate

Shoals, Northwestern Hawaiian Islands. Atoll Res. Bull., No. 150.

383 pp.

Atkinson, I. A. E. 1985. The spread of commensal species of Rattus to

oceanic islands and their effects on island avifaunas. Pp. 35-81.

In Conservation of island birds. ICBP Tech. Publ. No.3.

Cambridge, England.

Best, L. B. and K. L. Petersen. 1982. Effects of stage of the breeding

cycle on Sage Sparrow detectability. Auk 99: 788-791.

Boyce, M. S. and C. M. Perrins. 1987. Optimizing Great Tit clutch size

in a fluctuating environment. Ecology 68(1): 142-153.

Dennis, B., P. L. Munholland, and J. M. Scott. In press. Estimation of

growth and extinction parameters for endangered species. Ecol.

Monographs.

Ely, C. A. and R. C. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll Res. Bull. No. 171.

361 pp.

Eddinger, C. R. 1970. A study of the breeding behavior of four species

of Hawaiian honeycreepers (Drepanididae). Ph.D. dissertation, Univ.

of Hawaii, Honolulu. 212 pp.

Ely, C. A. and R. C. Clapp. 1973. The natural history of Laysan

Island, Northwestern Hawaiian Islands. Atoll Res. Bull., No. 171.

361 pp.

249

------Johnson, N. K., J. A. Marten, and C. J. Ralph. 1989. Genetic evidence

for the origin and relationships of Hawaiian honeycreepers (Aves:

Fringillidae). Condor 91: 379-396.

Loope, L. L., O. Hamann, and C. P. Stone. 1988. Comparative

conservation biology of oceanic archipelagoes: Hawaii and the

Galapagos. Bioscience 38: 272-282.

Morin, M. P. 1987. Laysan Finches drown as a result of marine debris.

'Elepaio 47: 107-108.

Marin, M. P. In prep. Laysan Finch nest characteristics, nest spacing,

and reproductive success in two vegetation types.

Morin, M. P. and S. Conant. 1990. Nest substrate variation between

native and introduced populations of Laysan Finches. Wilson Bull.

102: 591-604.

Murphy, M. T. 1986. Body size and condition, timing of breeding, and

aspects of egg production in Eastern Kingbirds. Auk 103: 465-476.

Newton, I. 1973. Finches. Taplinger Publishing Co., Inc., New York.

288 pp.

Pletschet, S. M. and J. F. Kelly. 1990. Breeding biology and nesting

success of Palila. Condor 92: 1012-1021.

Raikow, R. J. 1977. The origin and evolution of the Hawaiian

Honeycreepers (Drepanididae). Living Bird 15: 95-117. savidge, J. A. 1987. Extinction of an island forest avifauna by and

introduced snake. Ecology 68: 660-668.

Scott, J. M., C. B. Kepler, C. van Riper III, and S. I. Fefer. 1988.

Conservation of Hawaii's vanishing avifauna. Bioscience 38: 238-253.

250 Sibley, C. G. 1970. A comparative study of the egg-white proteins of

passerine birds. Peabody Mus. Nat. Hist. Bull. 32: 1-131.

Sibley, C. G. and J. E. Ahlquist. 1982. The relationships of the

Hawaiian Honeycreepers (Drepaninini) as indicated by DNA-DNA

hybridization. Auk 99: 130-140.

Sincock, J. L. and E. Kridler. 1977. The extinct and endangered

endemic birds of the Northwestern Hawaiian Islands. Unpubl. ms., U.

S. Fish and Wildlife Service, Honolulu. 111 pp.

Soule, M. E. 1980. Thresholds for survival: maintaining fitness and

evolutionary potential. Pp. 151-169. In Conservation biology: an

evolutionary-ecological perspective. M. E. Soule and B. A. Wilcox

(eds.). Sinauer Assoc., sunderland, Massachusetts.

Story, K. 1986. Pharaoh ants: a growing problem. Pest Control Tech.

(Feb.) :40-46. van Riper, C. III. 1978. The breeding biology of the Amakihi (Loxops

virens) and Palila (Psittirostra bailleui) on Mauna Kea, Hawaii.

Ph.D. thesis, Univ. of Hawaii, Honolulu. 166 pp.

Welty, J. C. 19B2. The life of birds. Saunders college Publishing,

New York. 754 pp.

251