Title: Temporal and social aspects of the foraging ecology of a piscivore, the osprey (Pandion haliaetus)
CITATION PDF VIEWER THUMBNAILS PAGE IMAGE ZOOMABLE
Full Citation
STANDARD VIEW MARC VIEW
Permanent Link: http://ufdc.ufl.edu/UF00099317/00001
 Material Information
Title: Temporal and social aspects of the foraging ecology of a piscivore, the osprey (Pandion haliaetus)
Physical Description: vii, 101 leaves : ill. ; 28 cm.
Language: English
Creator: Edwards, Thomas C
Copyright Date: 1987
 Subjects
Subject: Osprey   ( lcsh )
Birds -- Florida -- Newnans Lake   ( lcsh )
Birds -- Ecology -- Florida   ( lcsh )
Forest Resources and Conservation thesis Ph. D
Dissertations, Academic -- Forest Resources and Conservation -- UF
Genre: bibliography   ( marcgt )
non-fiction   ( marcgt )
 Notes
Statement of Responsibility: by Thomas C. Edwards, Jr.
Thesis: Thesis (Ph. D.)--University of Florida, 1987.
Bibliography: Bibliography: leaves 93-99.
General Note: Typescript.
General Note: Vita.
 Record Information
Bibliographic ID: UF00099317
Volume ID: VID00001
Source Institution: University of Florida
Holding Location: University of Florida
Rights Management: All rights reserved by the source institution and holding location.
Resource Identifier: alephbibnum - 000948866
notis - AER0973
oclc - 016902045

Downloads

This item has the following downloads:

temporalsocialas00edwa ( PDF )


Full Text












TEMPORAL AND SOCIAL ASPECTS OF THE FORAGING
ECOLOGY OF A PISCIVORE, THE OSPREY (Pandion haliaetus)








By

Thomas C. Edwards, Jr.


A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL
OF THE UNIVERSITY OF FLORIDA IN
PARTIAL FULFILLMENT OF THE REQUIREMENTS
FOR THE DEGREE OF DOCTOR OF PHILOSOPHY


UNIVERSITY OF FLORIDA

1987
















ACKNOWLEDGMENTS


Numerous individuals provided assistance throughout my tenure as a

graduate student in the Department of Wildlife and Ranges Sciences, and

to each I wish to express my gratitude and extend my thanks.

In particular, I thank my friend and graduate advisor, M. W.

Collopy, for his guidance through the ups and downs of graduate school.

His support and friendship transcended that required of a graduate

advisor.

Members of my graduate committee, J. F. Eisenberg, P. Feinsinger,

S. R. Humphrey, and H. F. Percival, graciously assisted me through all

phases of my research, from the initial development and discussion of

my research topic to suggestions for improvement and review of my

dissertation. Their open office doors and willingness to discuss more

than just science have been much appreciated.

Other faculty in the Department of Wildlife and Range Sciences

aided my development as an academician and a person. In particular, I

wish to thank W. Kitchens for first introducing me to conceptual

"popcorn farts," and H. F. Percival for his constant reminders that

there is more to life than graduate school. My research also benefited

form numerous discussions with J. Robinson.

My fellow graduate students, who always were willing to sharpen

their intellectual claws on my research, helped improve in that manner











peculiar to graduate students the quality of my dissertation research.

J. Wiechman first introduced me to the idiosyncrasies of

electrofishing, greasy wheel-bearings, and the "fish of the month"

club. R. Bennett taught me how to climb trees without shredding my

ankles, and I now understand why you do not look down. T. B. Murphy's

constant companionship sometimes was appreciated. Numerous

undergraduate and graduate students, too many to name, assisted during

electrofishing samples. To all I extend my thanks.

Primary financial support was provided through Mclntire Stennis

Project No. 2409. Logistical and other financial support was provided

by the Florida Cooperative Fish and Wildlife Research Unit, The Hawk

Mountain Sanctuary Association, and a Grant-in-Aid-of-Research from

Sigma Xi. Additional financial support was provided by my parents, who

always seemed to sense when a few extra dollars would come in handy.

Their recognition and understanding of the difficulties associated with

raising a family while simultaneously completing a degree has been a

great comfort.

Last, I thank my wife, Kathy, for her continuous support

throughout these sometime trying times. Her ability to provide for me

and our two children while I completed my degree was a far greater

accomplishment than mine.
















TABLE OF CONTENTS

Page

ACKNOWLEDGMENTS . . . . . . . . . . . . 1

ABSTRACT . . . . . . . . . . . . . vi

CHAPTERS

I INTRODUCTION AND OVERVIEW . . . . . . . . 1

General Background . . . . . . . . . 1
Study Organization . . . . . . . . . 3

II TEMPORAL VARIATION IN THE FORAGING ECOLOGY OF OSPREYS . 5

Introduction . . . . . . . . . . . 5
Methods . . . . . . . . . . . . 8
Study Area . . . . . . . . . . . 8
Evaluation of Prey Base . . . . . . ... 9
Osprey Foraging Behavior . . . . . . . 10
Analysis . . . . . . . . . . . 11
Results . . . . . . . . . . . . 13
Prey Base . . . . . . . . . . . 13
Prey Relative Abundance . . . . . . . 14
Prey Absolute Abundance . . . . . . . 17
Prey Patchiness . . . . . . . . . 20
Adult Foraging Patterns . . . . . . . . 25
Best-fit Model . . . . . . . . . . 26
Effect of Habitat . . . . . . . . 29
Effect of Sex . . . . . . . . . 31
Effect of Time . . . . . . . . . 36
Discussion . . . . . . . . . . . 39

III THE ONTOGENY OF PREY PREFERENCE:
EFFECT OF A VARIABLE RESOURCE BASE . . . . . 44

Introduction . . . . . . . . . 44
Methods . . ... . . . . . . . 46
Study Area and Evaluation of Prey Base . . . . 46
Fledgling Foraging Behavior . . . . . . . 46
Analysis . . . . . . . . . . 48
Results . . . . . . . . . . . . 49
Prey Base . . . . * * ....... . . 49
Fledgling Foraging Patterns . . . . . . . 55

iv










Model PS,PB,PT . . . . . . . . . 57
Model PB,ST . . . . . . . . . ... . 60
Fledgling Versus Adult Foraging Patterns . . . 63
Discussion . . . . . . . . . . . 67

IV SIBLING ENHANCED FORAGING IN OSPREYS . . . . . 72

Introduction . . . . . . . . . . . 72
Study Area and Methods . . . . . . . . 73
Results . . . . . . . . . . . . 75
Discussion . . . . . . . . . . . 83

V CONCLUSION AND SYNTHESIS . . . . . . . . 88

LITERATURE CITED . . . . . . . . . . . . 93

BIOGRAPHICAL SKETCH . . . . . . . . . . . 100
















Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy


TEMPORAL AND SOCIAL ASPECTS OF THE FORAGING
ECOLOGY OF A PISCIVORE, THE OSPREY (Pandion haliaetus)


By

Thomas C. Edwards, Jr.

May 1987


Chairman: Michael W. Collopy
Major Department: Forest Resources and Conservation
(Wildlife and Range Sciences)

I conducted a year-around field study on foraging behavior of a

resident population of ospreys (Pandion haliaetus) at Newnan's Lake,

Alachua County, Florida, from 1985-1986. I examined (i) foraging

behavior of adult ospreys in relation to temporal variation in

available prey, and (ii) temporal and social aspects of the ontogeny of

foraging. I also monitored dynamics of the fish resource base at the

study site.

Adult foraging behavior varied considerably over the 18-month

study period. Ospreys preferred sunfish (Lepomis spp.) from March-July

each year, but preferred shad (Dorosoma spp.) from September-March.

Bass (Micropterus salmoides; Morone saxtilis) were captured in

proportion to their abundance. Seasonal shifts from shad to sunfish

were strongly associated with increases in sunfish relative and










absolute abundances. Once sunfish abundance declined, ospreys switched

back to shad. Relative to sunfish, measures of shad abundance and

variability were more constant, suggesting that shad represent a more

stable resource base. Time lags between change in prey abundance and

shifts between prey types may be due to the inability of ospreys to

respond immediately to changes in the prey resource base.

Adults preferentially hunted in the littoral zone of the lake from

May-July each year. From August-April, ospreys hunted in pelagic

habitat, although preference for pelagic habitat was not statistically

significant. Use of littoral habitat was strongly associated with an

increase in the abundance and availability of sunfish.

Young ospreys initially captured fish in proportion to their

availability by species and size class, ignoring only the larger size

class of all fish species. Individual differences in preference

patterns existed between unrelated young. Siblings, however, hunted

together and had statistically similar prey preference patterns

throughout post-fledging. Use of fish resources and foraging mechanics

of birds with siblings also approached that of adults at a faster rate

than lone birds. Similar resource use and faster rates of learning

between siblings suggests post-fledging interactions may facilitate the

development of foraging skills.















CHAPTER I


INTRODUCTION AND OVERVIEW



General Background



Diet choice of predators has received considerable attention

within the framework of optimal foraging theory. A major

disappointment, however, has been the apparent lack of agreement

between model predictions and results, and the associated recognition

that early foraging models (MacArthur and Pianka 1966, Schoener 1971,

1974, Pulliam 1974, Charnov 1976a, b) may be overly simplistic. To

overcome criticisms of simplicity, numerous second generation models

that take into account such factors as nutrient requirements (Pulliam

1975), prey-recognition time (Hughes 1979), and prey-handling time

(McNair 1981) have been proposed. Unfortunately, tests of the precise

quantitative predictions of both first and second generation models

often provided only "qualitative" support regarding predictions of diet

choice (e.g., Werner and Hall 1974, Emlen and Emlen 1975, Davies 1977,

Pulliam 1980, Krebs et al. 1983).

A theme common to most foraging models is that a predator operates

within the framework of only a single foraging strategy set. That is,

one model is capable of describing most aspects of predator diet










2

choice. Consequently, deviations in diet choice from those predicted

frequently are interpreted as "mistakes" (Jaeger and Barnard 1981) or

as "sub-optimal" behavior (see Pyke 1984) in what appears to be an

attempt to fit contradictory results to a particular model.

Unfortunately, this perception ignores the possibility that

predators might utilize a suite of diet choice algorithms, switching

from one to the other in response to changes in the available resource

base (Glasser 1982, 1984, Glasser and Price 1982). Such a facultative

strategy, where the degree of prey discrimination and choice varies

with change in prey abundances, should be advantageous to predators

facing highly variable environments. In contrast, predators can

concentrate on particular subsets of the available prey base under

constant resource conditions. Here, a single model may adequately

explain diet choice.

Finally, a key component of foraging behavior and diet choice

studies is knowledge about temporal changes in the prey resource base.

For instance, a constantly changing resource base requires a high

degree of flexibility in predator foraging behavior. One important

component of this flexibility is the ability to incorporate information

from a changing environment into a new foraging strategy. Therefore, a

close coupling of information on prey dynamics with foraging behaviors

is needed to fully understand foraging behaviors. Such is the approach

I have undertaken here.










3

Study Organization



My study is organized into two sections comprising three chapters.

The first section (Chapter II) is an examination of temporal shifts in

prey preference of adult ospreys in relation to prey dynamics.

Temporal aspects of foraging (e.g., Doble and Eggers 1978) are an

important yet frequently neglected area of predator foraging

strategies. For example, predator diet choice may reflect use of

nutritionally important prey types prior to the breeding season when

mobilization of energy for reproduction is important. Alternatively,

diet choice may become increasingly specialized over time as resource

abundance increases and predators are able to be selective. This

potential for seasonal variation in prey preference often is ignored

and can be addressed only by year-around study of the same group of

predators.

The second section (Chapters III and IV) concerns the ontogeny of

foraging in ospreys. The post-fledging period, when naive young first

confront an environment from which they must learn to capture prey, is

a crucial time for avian predators. The strongest selective influence

on foraging performance likely occurs during this period (Zach and

Smith 1981), with learned foraging skills undoubtedly affecting

subsequent survival of the young (Kamil 1983). Study of the ontogeny

of foraging also has important management implications. Currently,

efforts are underway to reintroduce fledgling ospreys to areas from

which they have been extirpated. Consequently, identification of

factors affecting fledgling survivorship during the post-fledging












period could aid in the development of appropriate management

strategies.

Chapter III is an ecological approach to questions regarding the

ontogeny of foraging. Here, I examine temporal aspects of the

development of prey preference in young ospreys and, in particular, how

variability in the prey resource base affects the learning process.

Study of how naive young respond to change in their resource base while

developing foraging skills could provide insights into foraging

behavior not obtainable from study of adults (Kamil 1983).

In Chapter IV I examine behavioral aspects of the development of

foraging skills. One hypothesized method by which young increase

foraging skills is observational learning (Turner 1964). Here, naive

young use experienced conspecifics as role models, learning from others

the appropriate cues that lead to successful foraging bouts. One

benefit of such social interactions may be an increased rate of

learning due to the more frequent exposure to appropriate foraging

behaviors. If social interactions lead to increased rates of learning,

managers may wish to modify reintroduction procedures to realize the

benefits associated with sociality. Together, Chapters III and IV

provide an ecological and behavioral perspective of the ontogeny of

foraging behaviors.















CHAPTER II


TEMPORAL CHANGE IN PATTERNS OF PREY PREFERENCE IN OSPREYS



Introduction



Suggestions that predators forage in an optimal manner first were

articulated by Emlen (1966) and MacArthur and Pianka (1966), who each

developed explicit mathematical models predicting how predators should

forage under certain circumstances. The basic premise underlying these

and subsequent models on foraging behavior and diet choice is that

optimal foraging, as contrasted with random foraging, maximizes

Darwinian fitness (see reviews by Schoener 1971, Pyke et al. 1977, Pyke

1984).

One criticism of diet choice models is the rapidity with which

both the quantitative and qualitative predictions break down when

models are modified in a realistic manner. For instance, models based

solely on energetic requirements (MacArthur and Pianka 1966, Schoener

1971, Pulliam 1974, Werner and Hall 1974, Charnov 1976a) predict that

partial consumption of lower-ranked prey should not occur, yet

contradictory (e.g., Willson 1971, Goss-Custard 1977) as well as

supportive results (e.g., Smith and Follmer 1972, Zach and Falls 1978)

are common in the literature. Furthermore, the lack of mutually










6

exclusive predictions often makes it difficult to determine which model

best describes predator diet choice.

Unfortunately, many foraging studies also have concentrated on

foraging behaviors only during discrete time periods (e.g.,

non-breeding season, Craig 1978; winter versus summer, Baker and Baker

1973, Opdam 1975). One consequence of this approach is the potential

loss of information on temporal aspects of foraging; in particular, how

diet choice at time t is affected by prey and predator dynamics at time

t-1. For example, temporal variation in prey preference may result

from both changes in prey abundance and particular stages in predator

life histories. Preference for a particular prey type may be strong

when prey abundances are high and predators have the option to be

"choosey," but less strong as abundances decline and predators are

forced to consider alternate prey. Another example may be the need for

"nutrient-rich" prey (e.g., Pulliam 1975, Westoby 1974, 1978) prior to

the breeding season. Once the breeding season finishes, preference for

nutrient-rich prey may decrease. In both instances, failure to

recognize the relationships among study period, prey abundances and

predator life histories makes interpretation of results difficult.

Moreover, relationships such as these would be apparent only by

year-around study of the same population of predators.

Here I examine the influence of a variable resource base on

patterns of prey preference in a resident population of ospreys

(Pandion haliaetus) in north-central Florida. Ospreys are ideal study

organisms for examining temporal relationships and prey preference

because of their high visibility and lack of susceptibility to certain











external constraints (e.g., predation, Martindale 1982) that often

modify foraging behaviors. Moreover, ospreys are obligate piscivores

and consume a resource amenable to sampling. Consequently, prey

dynamics are relatively easy to monitor.

Several objectives were accomplished by my research. First, I

examined the effect of change in the underlying prey resource base on

patterns of prey preference in ospreys. In general, a predator's

foraging strategy, and hence its diet, depends on whether it occupies

an environment where the prey base is constant or variable over time

(Schoener 1971, 1974, Pulliam 1974). Under constant resource

conditions, predators should exhibit either an obligate specialist or

obligate generalist strategy. If prey abundance is consistently low, a

generalist strategy should prevail, while a specialist strategy is

advantageous if prey abundance is relatively high and constant.

A facultative strategy, the implicit strategy of most optimal

foraging models, should be used in environments where the prey base is

highly variable. Under variable resource conditions, prey preference

patterns of ospreys should indicate a switch from a generalist to

specialist strategy as fish abundance increases. Measures of

preference at low fish abundance levels should indicate use in relation

to availability (i.e., "random" use), while preference should be

independent of frequency at relatively high abundance levels. Thus, a

major goal of my research was to determine the extent to which ospreys

switch from a generalist to specialist foraging strategy as fish

abundances varied over time.











Second, I wished to examine the possible role of time lags in

response by ospreys to changes in prey availability. Because an

important component of foraging models is information on prey

availability, it has been argued that predators should continuously

sample from the environment as insurance against future change (Smith

and Sweatman 1974, Oster and Heinrich 1976). Once abundances change

and use of a particular prey type is no longer profitable, predators

should switch emphasis and concentrate on the particular prey type

currently profitable (Murdoch 1969, 1973, Murdoch and Oaten 1975).

Whether shifts in preference patterns of ospreys occur

simultaneously with change in the fish resource base, or exhibit a time

lag, may depend on whether fish abundances shift gradually or abruptly

over time. Abrupt shifts in abundance that force ospreys to switch

preference undoubtedly introduce a time lag, while gradual shifts in

abundance should allow ospreys to simultaneously shift preference.

Consequently, examination of the possible role of time lags requires

close coupling of the dynamics of the fish resource base with osprey

foraging behavior.


Methods



Study Area

My research was conducted on Newnan's Lake, Alachua County,

Florida, from March 1985-September 1986. Newnan's Lake, located 15 km

east of Gainesville, Florida, is a shallow (<3 m depth), 2400 ha

hyper-eutrophic lake (Shannon and Brezonik 1972) rimmed with










9

baldcypress (Taxodium distichum) and mixed hardwoods. Most of Newnan's

Lake is open water, although from May-October much of the lake

periphery is covered by hydrilla (Hydrilla verticillata), coontail

(Ceratophyllum spp.), and spatterdock (Nuphar luteum) vegetation.

Changes in the lake vegetation structure are due primarily to

fluctuation in water level, which varies from a 3 m depth in winter

months to a 1.5 m depth in summer. This variation in water depth

creates different lake habitats at different times of the year. During

winter months, when the lake is deepest, the lake is relatively open

and vegetation is absent. From May-October, however, decreased water

depth allows vegetation to establish roots in the underlying lake

substrate and cover up to 10% of the lake surface area. The creation

of two lake habitats during this period, a littoral and pelagic zone,

affects the underlying fish resource base and presents ospreys with two

different foraging habitats.



Evaluation of Prey Base

I used electrofishing to estimate the fish resource base available

to ospreys (see review by Reynolds 1983). This technique uses an

electrical charge to stun fish for capture and subsequent measurement

or determination of desired parameters, and is considered an effective

method by which abundances can be estimated. One advantage of

electrofishing from the standpoint of this study is that it samples

fish at the top of the water column. Fish thus sampled represent what

is available to ospreys, who cannot dive to great depths to capture

prey. While electrofishing efficiency is known to be size related










10

(Sullivan 1956, Reynolds and Simpson 1978) and hence biased against

small fish ("fingerlings," <5 cm), estimates of length frequencies of

larger fish are considered reasonably accurate (Reynolds 1983).

Because ospreys at Newnan's Lake typically caught fish in excess of 10

cm (Collopy 1985, this study), I did not consider this bias against

fingerlings to be a disadvantage.

Fish availability was estimated monthly throughout the 18-month

study period. I ran 12 15-min electrofishing transects, six each in

the littoral and pelagic zones, on two consecutive days. Capture

results were used to estimate the relative abundance of each fish

species. Trends in absolute abundance were based on the number

captured per 15-min transect. Fish parameters collected included fish

species, weight (g), and total length (cm). I excluded from analysis

unlikely prey for ospreys (e.g., bottom-dwelling brown bullhead,

Ictalurus nebulosus) because I felt their inclusion would bias results

by inflating preference measures for other prey types.



Osprey Foraging Behavior

I observed osprey foraging behavior from a boat anchored offshore

at locations that facilitated simultaneous observations on several

birds. Individual ospreys were chosen at random and watched until the

completion of a 15-min period or until a successful capture was made

and the fish type and size determined. I preferred this approach over

continuous observations due to difficulties associated with maintaining

observation of the same individual, and because it increased the number









11

of independent samples. Distinction between sexes were made on the

basis of plumage characteristics (MacNamara 1972).

I collected data during two 5-day periods, one before and one

after the 2-day interval during which fish were sampled. Every effort

was made to ensure continuity in the sampling scheme, although

inclement weather and logistical problems frequently forced minor

adjustments in the sampling design.

Identification of prey species captured was relatively

straightforward owing to distinctive fish silhouettes and the

anterior-posterior manner in which ospreys carried fish to minimize

wind resistance. I assigned prey captured to one of three 10 cm size

classes (10-20 cm, 20-30 cm, 30-40 cm) based on relative length of the

fish to the bird (see Poole 1982). The approximate location and lake

habitat type of each capture was noted on 7.5 min USGS topographic maps

of the lake.



Analysis

I used log-linear analysis to develop statistical models best

describing use of the available fish resource base over time by

ospreys. Log-linear analysis is a procedure similar to

analysis-of-variance for use on multidimensional categorical data

(e.g., habitat type, prey size class) (Bishop et al. 1975, Feinberg

1980, Agresti 1984). The procedure examines models composed of all

factors and their interactions in a hierarchical fashion; that is,

whenever models containing higher order interactions are considered,

lower order effects composed of the same factors also must be included









12

in the model. The presence of significant interactions between factors

of any model suggests that that particular model does not fit the data.

Non-significant interactions for any combination of factors suggests

each factor is independent and, therefore, represents a model that fits

the data.

Hereafter, factor refers to the variables of interest, level to

categories within each factor, and cell to the intersection of >2

levels. Factors analyzed included prey type, prey size class, time

(month), habitat (littoral or pelagic), and osprey sex.

Model cells representing prey captured by each individual osprey

were weighted by the relative abundance of each fish species and size

class prior to analysis. Weighting standardizes the cell expected

value to its estimated frequency in the environment. For example, fish

species comprising 10% and 90% of the resource base, respectively,

should not have identical (i.e., 0.5) expected capture probabilities if

ospreys are foraging at random. Instead, capture data must be

standardized to the relative probability of encounter of each fish

species (here, 0.1 and 0.9, respectively).

Standardized lambda estimates were calculated for each cell in the

model. Basically, lambda estimates are a measure of the difference

between expected and observed values (i.e., standardized cell

residuals) that can be used as a measure of preference for a particular

fish species and size class. Lambda estimates have an asymptotic

standard normal distribution and can be compared to normal z-scores for

significance testing. Here, I considered significant positive lambda

estimates to indicate "preference" for a particular fish resource type.









13

Non-significant lambda estimates indicated random use. Comparisons of

lambda estimates and construction of confidence intervals were made

using formulas provided by Manly (1974) and Heisey (1985).

I first evaluated all possible models using the likelihood ratio

statistic, G 2, eliminating those with significant P-values. When

several models of increasing complexity (i.e., more factors) fit the

data, conditional tests (Agresti 1984:57-58) were used to determine the

best fit model. This approach subtracts G2-values and degrees of

freedom of more complex from less complex models to determine whether

the more complex model adds any additional information. A significant

G2-value suggests that the more complex model adds additional

information and should be considered. Once a "best-fit" model was

chosen, I calculated cell lambda estimates as described above.

The level of significance for all analyses was 0.05 unless

otherwise noted. All analyses were performed using procedures found in

Biomedical Computer Programs (Dixon 1985: BMDP-4F), Statistical

Analysis Systems (1982: CATMOD), and Statistical Package for the Social

Sciences (1986: LOGLINEAR) analytical guides.


Results



Prey Base

Fish data were collapsed into three categories representing bass

(largemouth bass, Micropterus salmoides; striped bass Morone saxtilis),

sunfish (warmouth, Lepomis gulosus; bluegill, L. macrochirus; redear

sunfish, L. microlophus), and shad (gizzard shad, Dorosoma cepedianum;









14

threadfin shad, D. petenense). Prey species were categorized for

several reasons. First, the fish species placed together are

behaviorally and ecologically similar. For instance, all three sunfish

interbreed (Breder and Rosen 1966:439-440, Childers 1967), as do both

species of shad (Minckley and Krumholz 1960). Second, food and

foraging habits of the categorized species are similar (references in

Breder and Rosen 1966, Carlander 1969, 1977). Last, and most

important, body forms of grouped fish species are similar, and it is

unlikely that ospreys are capable of distinguishing minute differences

(e.g., presence of red dot on lateral surface of redear sunfish)

between species in the absence of gross behavioral differences.

Prey relative abundance. Prey relative abundances varied across

species groups, size classes, and time. Major differences existed

primarily between lake habitats, with bass not present in the pelagic

zone and sunfish found there only seasonally. Furthermore, overall

abundance of all fish species groups was higher in the littoral than in

the pelagic zone.

Bass were not found in the pelagic zone (Fig. 2-1). In the

littoral zone, abundance was fairly constant within each size class,

but no temporal pattern of change was obvious. Not unexpectedly,

differences in availability among bass size classes reflect size class

rankings, with each smaller size class more prevalent than its next

larger size class.

Shad were found in approximately the same proportion in pelagic

and littoral habitats (Fig. 2-1). A major difference, however, was the

shift in the most common size class between lake habitats. Smaller













Figure 2-1. Relative abundance of each fish species group in littoral and pelagic habitat.




















HABITAT TYPE: LITTORAL

0.2 BASS


0 Io\ -oo

w0..
I I I I I I I I I I I I I I I I I


z

A aH ,oo /,0-

0." A. A ..

0 . A A . ...--, ....?..... ... .


I
'-

0.5



0.
0



0 0.1 -
0-
o 0.2 -


0-


SUNFISH





I a a A


S / \

o. o V *** o..-**

...o....o... ...o....o ..... o- .. o .. ..o....o .o. o...y .... o
S i I I I I I I I I I I I I
M A M J J A S O N D J F M A M J J A

1985 1986


HABITAT TYPE: PELAGIC

BASS NOT PRESENT





I I i I I I I I I I i i i



SHAD




on .. ., 0, o. --o ..- ..- ..- ..o ..-...




SUNFISH



o---o 10-20 cm

A-- 20-30 cm

0.........0 30-40 cm





^ ..,t A . ; o 8 ..O...

M A M J J A S N D J F M A M J J A

1985 1988










17

shad (10-20 cm) were more prevalent in the littoral zone, while 20-30

cm sized shad were more common in pelagic habitat. This difference

probably reflects the tendency of smaller-sized individuals to remain

in shallower waters until they are sufficiently large to survive and

compete in deeper waters. The numbers of shad in the 30-40 cm range

were consistently low in both lake habitats. Unlike bass, the shad

population appeared to cycle, with high and low abundances occurring in

the winter and summer months, respectively.

Sunfish exhibited patterns different from bass and shad. They

were present year-around in the littoral zone but only seasonally in

the pelagic zone (Fig. 2-1). In the littoral habitat, 20-30 cm sized

sunfish exhibited a cyclic pattern, peaking between May and August. No

such obvious pattern was apparent for 10-20 cm size class, although

there appeared to be a general decrease in abundance in this size class

throughout the study period. Whether this represents a longer cycle is

not known. Sunfish 30-40 cm long were fairly constant in abundance but

represented a small proportion of the available fish overall. Sunfish

also exhibited two brief, distinct movements into the pelagic zone.

Prey absolute abundance. Patterns of change in absolute abundance

were, in general, similar to those exhibited for relative abundance.

Estimates of relative and absolute abundance of bass revealed no

apparent differences for any of the three size classes (Fig. 2-2).

Similarly, patterns of absolute abundance for shad in littoral habitat

were similar to those of relative abundance. In the pelagic zone,

however, patterns of absolute abundance of 10-20 cm and 20-30 cm sized

shad differed from relative abundance (Fig. 2-3). Instead of










Figure 2-2. Mean number (+ 1 SD) per 15-min sample transect (n = 6) of
each fish species group and size class in littoral habitat.


























NO. PER 15-MIN TRANSECT


- -






-- a
-0 0 o




o -0-- La -


-0---0- -0-
(to
















-0 0- -0--




0- -0- -0.





-0- -0- -0




0 0 -0 -0-
-0 -0--- -0--



o^ a -o--


0 0 a











-0- -o-- -0-


4---- 0-
4 --o- --













^ U M
----







--0- o- -





4-0- -0 -0-(-








I- I


o0 0

-- 3 -- 3 -0- 3


0 0 0








-- --0- --






0- -0- -0-




-0- 0- o-












-0- .- -0-











- O- a -o- 0

0 o- a -o- a
-0- -0-- 3 ~


-



>
O -










C-

0-


2-




. a-







-






C -


I




"r
-4


-4
-C








I-
.-4

0


r-










20

fluctuating, numbers of shad in these two size classes were fairly

constant over time. Numbers of shad in the pelagic zone also were more

constant than numbers in the littoral zone. Patterns of change in

absolute abundance of sunfish were similar to those of relative

abundance in both littoral (Fig. 2-2) and pelagic (Fig. 2-3) habitats.

Distinct cycles in fish numbers in both lake habitat types, and

when habitats are combined, are apparent when estimates of absolute

abundance are summed across prey groups and size classes (Fig. 2-4).

In general, fluctuations in prey absolute abundance were greater in the

littoral than in the pelagic zone. Thus, the pelagic zone represented

a more constant habitat with respect to prey numbers than did the

littoral zone.

Prey patchiness. Variation or "patchiness" in distribution of

each fish species and size class is shown in Figures 2-2 and 2-3. To

determine whether variation was constant through time for each fish

group and size class, Bartlett's test for homogeneity (Sokal and Rohlf

1981:403-412) was applied to the monthly electrofishing sample

variances. With the exception of 30-40 cm sized shad

(Chi-square = 18.21, df = 17, P > 0.05), all fish species groups and

size class combinations in the littoral habitat exhibited significant

differences in variance over time (Chi-square > 32.53 for each

combination, df = 17, P < 0.05). In contrast, variances in 10-20 cm

and 20-30 cm sized shad in the pelagic zone were constant over time

(Chi-square < 22.67 for each size class, df = 17, P > 0.05). Samples

for sunfish and for 30-40 cm sized shad were not sufficiently

continuous for comparison of variances.










Figure 2-3. Mean number (+ 1 SD) per 15-min sample transect (n = 6)
of each fish species group and size class in pelagic
habitat. Bass were not present in pelagic habitat.















NO. PER 15-MIN TRANSECT








-- .-o-- -o



coa




> -- .-o- -o-- o-
01 > --o- -o-0 --4- 0- --






S- -o-I -o- -
- -o- -0-- --0- -0o--





S00 -I- -- 0-
> -0- -0- -0o- -o0- -0o- -
01 -0- -0- -0-- -0- "O





0 -0- -- --




"D "











- o 0-o- --o o -o-- -0-













Figure 2-4. Mean number of fish captured per 15-min sample transect (n = 6) in littoral
and pelagic habitat, and both habitats combined. Data are summed across all
fish species groups and size classes.












o0.... Total

D--o Pelagic

A-A Littoral


A


s-
50 -

C0)
z

oC 40 -
I-
z
I
U) 30
Cr
UJ
0.
- 20 -
0
O
U-
Z 10 -
z
w


/
/ ,'


I I I I I I I I I I I I I I I I I I
M A M J J A S O N D J F M A M J J A

1985 1986










25

One problem with the use of Bartlett's test as a measure of

heterogeneity in variance of each fish group and size class over time

is its sensitivity to non-normal distributions. Because the underlying

distribution of the fish groups more closely follows a Poisson than a

normal distribution, these differences may reflect non-normality of

data rather than differences in variances. Nonetheless, use of

Bartlett's test here provides some idea whether variation in the

abundance of each fish group and size class is constant or variable

over time.

In general, differences in prey abundances between lake habitat

types presents ospreys with two different foraging environments. Prey

from all three fish groups were present year-round in the littoral

zone, although abundances varied. In contrast, bass were not found in

pelagic habitat and sunfish were found there only seasonally. Shad

were found in both lake habitats, but were more variable in the

littoral than in the pelagic zone.



Adult Foraging Patterns

A total of 2823 successful captures (roughly 15 per observation

day) were observed over the 18-month study period. Because of small

sample sizes relative to the number of classifying variables, I was

forced to collapse data from each observation day into monthly periods

centered about each electrofishing sample. Collapsing daily data into

monthly periods also was required to prevent generation of cell

expected values less than 1.0.










26

Several factors also forced me to combine individual foraging data

and examine foraging behavior at a population rather than individual

level. First, I was unable to capture and color-mark adults. Instead,

I determined the identity of particular birds by observing their return

to individual nesting trees and, occasionally, their unique plumage

characteristics. While this allowed for identification of individuals

during the breeding season, its accuracy as a means of identification

during the non-breeding season is not known.

Preliminary analysis of data on individually recognized birds

(roughly 35% of the entire data set) revealed some differences among

individuals. In general, however, these differences were caused by <3

of the 21 known birds. The remaining birds were similar statistically.

I therefore concluded that my collection and analysis of data at a

population level adequately represented individual foraging behaviors.

Even though Chesson (1984) suggests foraging data must ultimately be

analyzed at the individual level, my analysis of data at a population

level should, at worst, widen confidence intervals about my measures of

preference. Second, as described above, I had to combine data to avoid

generating expected cell values less than 1.0.



Best-fit Model

Tests of partial and marginal associations were used to determine

which factors and interactions were necessary in the first-pass models.

Non-significant partial and marginal associations suggests a particular

interaction between two factors is not needed in the model.

Interactions having both significant partial and marginal associations










27

are considered necessary, while those having only one significant

association are considered questionable (Brown 1976).

Hereafter, shorthand notation will be used to describe the

relationships among the factors, where F represents the available fish

species groups, C the fish size classes, S osprey sex, H the lake

habitat type, and T the time interval. For example, the term FCT means

that prey types, size class, and time have statistically significant

associations with one another. Biologically, the term states that use

of the available fish species groups and size classes varied over time.

Tests of partial and marginal associations revealed that no

four-factor interactions had significant marginal or partial

associations. Consequently, all four-factor terms were excluded from

consideration. Four of 10 possible three-factor interactions had

significant partial and marginal associations, while 3 of 10 had either

a significant partial or marginal association (Table 2-1). All

two-factor interaction terms had at least one significant test.

Because of the high number of possible models and their complexity, I

decided to consider only those terms having both significant partial

and marginal associations. Terms retained for possible models included

FCT, CHS, CHT, and CST. Because of the hierarchial nature of

log-linear models, all possible two-factor terms with significant

partial and marginal associations were contained within the

three-factor terms and were automatically considered.

The model FCT,CHS,CHT,CST represents the most complex, but not

necessarily the most parsimonious, model. Other models less complex,

but containing the necessary significant interaction terms, also were












Table 2-1. Tests of partial and marginal associations for the model
factors prey type (F), prey size class (C), habitat (H),
osprey sex (S), and time (T). No four-factor interaction
terms had significant partial or marginal associations and
were not considered.


Effect Partial Association Marginal Association

df G2 P-value df G2 P-value

Two-factor terms


Three-factor terms


137.3
0.2
1.6
176.7
187.2
58.8
120.5
42.3
69.9
743.5



0.6
4.5
146.7

15.9
37.6
21.5
54.9
48.9
17.9


<0.001
0.738
0.444
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001



0.438
0.341
<0.001

0.069
0.307
<0.001
0.002
0.047
0.390


47.3
3.9
10.9
192.5
57.8
28.1
121.0
65.8
109.8
727.0



2.8
3.4
127.9

10.7
45.3
51.8
58.9
59.6
20.2


<0.001
0.049
0.005
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001
<0.001



0.091
0.492
<0.001

0.296
0.093
<0.001
0.008
0.004
0.265


a Pattern of structural zeros causes a negative df.


See methods.









29

evaluated (Table 2-2). Five of the 9 possible models had

non-significant P-values, suggesting each of these models adequately

fit the data. Conditional tests among these models revealed that

ST,FCT,HS,CS,CHT provided the best fit. This model states that prey

choice among adult ospreys varied simultaneously with prey size class

and time (FCT), that size classes used varied simultaneously over

habitat type and time (CHT), and that differences between the sexes

existed over time, between habitat types, and among prey size classes

(ST, HS, and CS, respectively). In general, the model terms represent

differences in prey use over time, between sexes, and between habitat

type. Each set of differences will be considered individually.

Effect of habitat. Foraging habitat was strongly associated with

osprey sex (HS, see below), and with prey size class and time (CHT).

The term CHT states that temporal differences in use of the prey size

classes existed independent of prey type, and that they were related to

habitat. To properly interpret CHT requires that each term be

evaluated in light of the other two terms.

Examination of CH revealed no relationship between foraging

habitat and each prey size class (lambda < 0.127 and P > 0.05 for all

three size classes). This suggests the association between CHT is due

primarily to interactions between CT and HT. Examination of lambda

estimates for CT revealed no apparent patterns. Only 2 of 54 possible

time and prey size class interactions deviated from random, suggesting

these values represent possible outliers that overly influenced the G2

statistic for this term.












Table 2-2. Summary of possible
ospreys in 1985-86.
associated with one
described in text.


models descibing foraging by adult
Factors not separated by commas are
another and should be interpreted as


Model df G P-value


CST,FCT,CHS,CHT 175 156.2 0.842
CST,FCT,HS,CHT 177 180.8 0.408
CST,FCT,CHS,HT 204 211.0 0.353
ST,FCT,CHS,CHT 209 216.5 0.347
ST,FCT,HS,CS,CHT 211 236.8 0.107
FCT,CHS,CHT 226 924.7 <0.001
CST,FT,FC,CHS,CHT 236 302.4 0.002
ST,FCT,CHS,HT 238 275.7 0.047
ST,FT,FC,CHS,CHT 270 362.8 <0.001










31

Temporal differences in use of habitat, however, were very

apparent (Fig. 2-5). Ospreys hunted preferentially in the littoral

zone during the brood-rearing and post-fledging months each year.

Negative lambda estimates indicate ospreys hunted more frequently in

the pelagic zone during the non-breeding season (August-April), but its

use was not statistically significant.

Effect of sex. Lack of an interaction between sex and prey types

means that preference for particular prey types was constant between

sexes. Instead, differences between sexes were associated with

habitat, prey size classes, and time.

Sexual differences in preferred foraging habitat were pronounced

and constant over time. Males tended to forage in the littoral zone

while females exhibited preferential use of the pelagic habitat

(lambda = 0.186, SE = 0.027, P < 0.001).

Differences also were apparent in the use of the fish size

classes. Males exhibited preference for the 10-20 cm size class

(lambda = 0.280, SE = 0.044, P < 0.001) while ignoring the 30-40 cm

size class (lambda = -0.271, SE = 0.074, P < 0.001). Females exhibited

the opposite pattern, concentrating on prey in the 30-40 cm size class

and ignoring 10-20 cm sized prey types. Neither sex exhibited

preference for 20-30 cm sized prey, using prey of this size class

equally and at random (lambda = 0.009, SE = 0.043, P > 0.05). These

differences were constant across all fish species groups and time.

Temporal differences in capture frequencies also existed between

the sexes (Fig. 2-6). Males hunted relatively more frequently than

females prior to the start of each breeding season (February-March)













Figure 2-5. Change in relation to time of preferred hunting habitat of adult ospreys. Histograms
with crosshatching represent statistical preference (P < 0.05) for either littoral or
pelagic habitat. Vertical lines are standard errors.




















Littoral


Pelagic


I I I I I I I I I I I I I I I I I I
MA M J JASON D J F MA M J J A


1985


1986


LU 1.0



< 1- .
wJ 1.0













Figure 2-6. Change in relation to time of hunting effort by male and female ospreys. Lambda
estimates are standardized to sampling effort and number of each sex present.
Cross-hatched areas represent significant differences (P < 0.05) in hunting effort.
Asterisks denote incubation, when females were not hunting. Vertical lines are
standard errors.




















jl 1.0


10


w 1.0


Males


I I I I I I I I I I I I I I I I I I
M A M J J A S 0 N D J F M A M J J A

1985 1986










36

while females hunted more frequently from June to August. Because

results were standardized to sampling effort and number of each sex

present, these differences represent true differences in hunting effort

over time.

Effect of time. Use of the available fish species groups and size

classes was strongly affected by time (Fig. 2-7). Sexes did not differ

in their use of the available fish species groups, nor was there an

association between habitat and use of prey. Ospreys did not exhibit

consistent preference for bass, essentially capturing bass at random

(Fig. 2-7a).

With few exceptions, preference for shad occurred during the

non-breeding season (Fig. 2-7b). During this period, osprey exhibited

statistical preference for shad in the 20-30 cm size class. Unlike in

sunfish, however, no one size class was consistently under-utilized.

Instead, there existed positive, non-significant preferences for all

size classes over time.

Use of sunfish exhibited a cyclic pattern over the 18-month study

period (Fig. 2-7c). Peaks occurred in May-August 1985 and in

March-July 1986. Examination of the 8 time periods when statistical

preference was expressed revealed that adults preferentially

concentrated on sunfish in the 20-30 cm size class. In two of the

periods they also exhibited preference for 10-20 cm sized sunfish.

Sunfish in the 30-40 cm size class were consistently under-utilized,

although none of the lambda estimates was statistically significant.













Figure 2-7. Change in relation to time of preference for each fish species group (A = bass, B = shad,
C = sunfish) and size class by adult ospreys. Each histogram triplet represents, left to right,
10-20 cm, 20-30 cm, and 30-40 cm sized fish. Histograms with cross-hatching represent
significant preference (P < 0.05) or under-utilization of a particular fish species group and
size class. Vertical lines are standard errors. Asterisks represent lambda estimates that
could not be estimated reliably.























A BASS


tU
I- B SHAD











1.0
F-














-1.0 -


M A M J J A S O N O

198865


J F M A M J J A

1986









39

Discussion



Shifts in diet choice over time suggest ospreys are switching prey

in the manner described by Murdoch (1969) and Murdoch and Oaten (1975).

Ospreys exhibited distinct patterns of preference, shifting preference

from sunfish to shad then back to sunfish. These shifts were strongly

related to concurrent shifts in sunfish abundance, with preference for

sunfish exhibited during the period they were most abundant. Shad, in

contrast, were relatively constant throughout the duration of my study.

Ospreys exhibited preference for shad only during periods of low

abundance of sunfish.

Associated with the shift from sunfish to shad was a shift in

foraging habitat. Ospreys preferentially hunted in the littoral zone

during peak abundances of sunfish and tended to use pelagic habitat

when concentrating on shad. It is important, then, to distinguish

between prey switching between habitats and within habitats. The

former is considered a consequence of predator aggregation (Krebs et

al. 1983), while the latter is a response to changes in the proportions

of available prey. The switch between foraging habitats documented

here raises several questions. First, why did ospreys switch to

pelagic habitat to forage on shad rather than continue to forage in the

littoral zone, particularly since shad were present in littoral

habitat?

Switching between different prey types in different habitats is

considered a function of both predator and prey densities (Royama 1970,

Murdoch et al. 1975). Predators should be expected to switch foraging












habitat when profitability decreases. The applicability of aggregation

to switching behavior exhibited by ospreys is, however, questionable.

Osprey numbers were not increasing in the time periods before the

habitat switch, nor is it likely osprey foraging had a significant

impact on the fish resource base. Rather, the switch to pelagic

habitat and use of shad appears primarily related to decreases in the

abundance of sunfish in littoral habitat.

A second question revolves around attributes of shad abundance

that encouraged the switch from littoral to pelagic habitat. Within

habitats, predators are thought to be sensitive to change in prey

absolute abundance (e.g., Hughes 1979). If so, then ospreys should

have switched to shad in the littoral zone since their absolute

abundance was increasing as sunfish decreased in abundance. Instead,

the shift to pelagic habitat suggests some attribute of shad abundance

was more favorable in the pelagic zone. One difference may be

potential encounter rates. Several authors have suggested predators

may be more sensitive to prey variance than mean values (e.g., Caraco

1980, Real 1980). Although mean absolute abundances of shad were

roughly similar in both habitats, variance measures were higher in

littoral than in pelagic habitat. Consequently, ospreys may have had

higher encounter rates and thus more successful foraging bouts in

pelagic habitat.

Instead of shifting only along a single continuum represented by

prey, shifts also could occur along a second continuum incorporating

habitat differences. Switches along habitat lines most likely would be

a function of the frequency and magnitude of change in the respective









41

prey resource bases (Janetos and Cole 1981). For example, the switch

to pelagic habitat by ospreys after sunfish abundance decreased could

represent a shift from a specialist strategy in littoral habitat to one

in pelagic habitat. Based on the patterns shown here, shad in pelagic

habitat represent a relatively constant resource base. Ospreys

switched to littoral habitat when the abundance of sunfish increased to

a point where numbers alone made their capture advantageous. Thus,

rather than shifting to a more generalist strategy, predators may

remain specialists by switching to alternate habitats provided these

habitats contain high prey abundances.

Preference patterns for the three prey types suggested ospreys

were specialists when overall fish abundance was high and fairly

general in diet choice when abundance was low. From May-September 1985

and June-August 1986 ospreys preferentially captured sunfish in the

10-20 and 20-30 cm size classes. Preference for sunfish, however, did

not develop until overall prey abundance peaked. Moreover, continued

preference for sunfish even when overall abundance was decreasing

suggests there existed a time lag from when fish abundances changed

until ospreys modified their diet. Instead of responding immediately

to change in sunfish abundance, ospreys continued to use sunfish for a

short time period.

Temporal lags in change of preference such as the one exhibited by

this population of ospreys are likely in predators foraging in variable

environments, and may be due to delay in the recognition by the

predator that previously ignored prey types now constitute profitable

food. This recognition may require the formation of new search images










42

that increase the predator's ability to recognize and capture different

prey types (Mueller 1971, Pietrewicz and Kamil 1979, McNair 1981).

Consequently, predators may continue to capture less profitable prey

types until a new search image is formed. Similarly, the time required

to develop a new search image could explain why ospreys did not

immediately respond to increases in the abundance of sunfish. Instead,

change in preference was gradual. Thus, rather than the abrupt shift

in preference predicted by some foraging models (e.g., "two-armed

bandit" model, Krebs et al. 1978, see also Hughes 1979), change in

preference could occur gradually. Lag in preference could simply

reflect the time required to develop a new set of search images

appropriate for a different resource base.

New search images probably develop during "intermediate" periods

of prey abundance, during which time prey preference should be

frequency dependent (Murdoch 1969, Fullick and Greenwood 1979, but see

Horsley et al. 1979). Here, if absolute abundances are summed across

both habitat types, the most abundant fish species group during these

"intermediate" periods is shad. Unfortunately, similarity in fish

absolute and relative abundance patterns makes it difficult to

determine whether preference patterns of osprey were frequency

dependent or proportional to availability. In general, however, osprey

preference patterns indicated a switch from preference for a subset of

the available prey (sunfish) to generality in diet.

Strict concordance between osprey preference patterns and fish

abundances seems unlikely unless changes in abundance are gradual.

Although the duration of my study was only 18 months, the apparent









43

regularity in peaks of fish abundance suggests that fish abundance

cycles regularly at Newnan's Lake. Whether the length of a time lag is

a function of cycle length in prey is not known. The ability of

facultative foragers to exhibit strict concordance between preference

and abundance may depend on the periodicity of prey (e.g., Craig et al.

1979). Gradual change would allow sufficient opportunity for predators

regularly to sample and respond to change. In contrast, rapid change

in abundance increases the potential for introduction of a time lag in

change of osprey preference patterns.














CHAPTER III


THE ONTOGENY OF PREY PREFERENCE:
EFFECT OF A VARIABLE RESOURCE BASE



Introduction



Avian foraging studies tend to ignore the post-fledging period

during which foraging skills are developed by young, and instead

concentrate on behavior exhibited by adults. The strongest selective

influence on foraging performance, however, may occur during the

post-fledging period (Zach and Smith 1981). Skills learned during this

period undoubtedly affect survival of the young and may influence

foraging performance as an adult (Kamil 1983).

Improvement in foraging skills over time has been demonstrated for

a variety of bird species, including brown pelicans, Pelecanus

occidentalis (Orians 1969), starlings, Sturnus vulgaris (Stevens 1985),

herring gulls, Larus argentatus (Verbeek 1977), glaucous-winged gulls,

L. glaucescens (Searcy 1978), and royal terns, Sterna maxima (Buckley

and Buckley 1974). Missing from many of these studies, however, is

knowledge of the prey dynamics during the period young are first

sampling from an environment unfamiliar to them. Variability in the

environment undoubtedly affects the learning process in young birds, and

information on the dynamics of the prey base during this learning period

is required to understand fully how foraging skills are developed.
44










45

Here I examine the ontogeny of foraging in ospreys (Pandion

haliaetus), exploring the relationship between development of foraging

skills and dynamics of the fish resource base. First, I wished to

determine how prey dynamics during the post-fledging period affected

temporal aspects of the ontogeny of foraging. For example, does

preference for particular prey types develop at faster rates if the prey

resource base is relatively stable? Or, in contrast, does a highly

variable environment extend the learning period by making it difficult

for naive young to determine which cues are most relevant to foraging

decisions? In addition, I wished to compare young from different years

to determine whether foraging skills develop in a similar fashion if the

temporal sequence of variation in the fish resource differs between

years. Specifically, is a model that adequately explains the ontogeny

of foraging in one year applicable in the next?

Second, I characterized use of the available fish resources by

young birds, examining how use of available prey resources varied in

response both to fluctuation in fish availability and to the

continuously improving skills of young. This implies a two-factor

interaction affecting learning. As a bird improves its foraging skills,

perhaps learning to concentrate on a particular prey type, the

availability of that resource may suddenly change. Cues that previously

led to successful foraging bouts may no longer be relevant. Lastly, as

an indicator of how young forage, I compared their use of resources with

adults who simultaneously are sampling from the same resource base.









46

Methods



Study Area and Evaluation of Prey Base

I conducted my research on Newnan's Lake, Alachua County, Florida,

during April-September 1985 and 1986. Newnan's Lake, located 15 km east

of Gainesville, Florida, is a 2400 ha hyper-eutrophic lake (Shannon and

Brezonik 1972) rimmed with baldcypress (Taxodium distichum) and mixed

hardwoods. Most of Newnan's Lake is open water, although from May to

October much of the lake periphery is covered by coontail (Ceratophyllum

spp.) and spatterdock (Nuphar luteum) vegetation.

I used electrofishing to estimate the fish resource base available

to ospreys (see review by Reynolds 1983). This technique uses an

electrical charge to stun fish for capture and subsequent measurement or

determination of desired parameters, and is considered an effective

method by which fish abundances can be estimated.

I ran six 15-min electrofishing transects on two consecutive days

each month to estimate fish availability. The relative abundance of

each fish species was estimated from the capture data, while an index of

absolute abundance was calculated from the number of fish captured per

15-min transect. Fish parameters collected included fish species,

weight (g), and total length (cm).



Fledgling Foraging Behavior

To aid in identification of young, I climbed suitable nest trees

and marked young ospreys with unique color band combinations prior to

fledging. I observed fledgling foraging behavior from a boat anchored










47

offshore at locations that facilitated simultaneous observations on

several birds. Individual fledglings were chosen at random and watched

until the completion of a 15-min period or until a successful capture

was made and the fish type and size determined. I preferred this

approach over continuous observations due to difficulties associated

with maintaining observation of the same individual, and because it

increased the number of independent samples. Instances where either

individual or prey type and size class could not be determined were

excluded from analysis.

I collected data on osprey foraging behavior during a 10-day period

before and one after the 2-day interval during which fish were sampled.

Data on prey species and size captured were summarized by individual.

To allow comparisons among individuals, I standardized individual

foraging data to days since fledging and placed results in 30, 60, 90,

120, and 150-day categories. Each 30-day category was centered around

an electrofishing sample. I assumed that variation in fish availability

was negligible between the daily sampling periods, and that the number

of prey consumed was small relative to the number available.

Identification of prey species captured was relatively

straightforward owing to distinctive fish silhouettes and the

anterior-posterior manner in which osprey carry fish. I placed prey

captured into one of three 10 cm size classes (10-20, 20-30, 30-40)

based on relative length of the fish to the bird (see Poole 1982). Size

classes of 10 cm represent the best level of resolution I felt capable

of identifying.












Analysis

I used log-linear analysis (Agresti 1984) to develop statistical

models best describing use of the available fish resource base by

ospreys. Log-linear analysis is a procedure similar to

analysis-of-variance for use on multidimensional categorical data (e.g.,

prey size class, age class) (Bishop et al. 1975, Feinberg 1980).

Typically, the simplest model that fits the data is considered the most

parsimonious. When several models of increasing complexity fit the

data, I used conditional tests (Agresti 1984:57-58) to determine the

best fit model.

Model cells representing prey captured by each individual osprey

were weighted by the relative abundance of each fish species and size

class prior to analysis. Weighting standardizes the cell expected value

to its estimated frequency in the environment. For example, fish

species comprising 10% and 90% of the resource base, respectively,

should not have identical (i.e., 0.5) expected capture probabilities.

Instead, capture data needs to be standardized to the relative

probability of encounter of each fish species (here, 0.1 and 0.9,

respectively).

Standardized lambda estimates were calculated for each cell in the

model. Basically, lambda estimates represent the difference between

expected and observed values (i.e., cell residuals), and they can be

used as a measure of preference for a particular fish species and size

class. Lambda estimates have an asymptotic standard normal distribution

and can be compared to normal z-scores for significance testing. Here,

I considered significant positive lambda estimates to indicate











"preference" for a particular fish resource type. Non-significant

lambda estimates indicated random use. Comparisons of lambda estimates

and construction of confidence intervals were made using formulas

provided by Manly (1974) and Heisey (1985).

The level of significance for all analyses was 0.05 unless

otherwise noted. All analyses were performed using procedures found in

Biomedical Computer Programs (Dixon 1985: BMDP-4F), Statistical Analysis

Systems (1982: CATMOD), and Statistical Package for the Social Sciences

(1986: LOGLINEAR) analytical guides.



Results

Prey Base

Fish data were collapsed into three categories representing bass

(largemouth bass, Micropterus salmoides; striped bass Morone saxtilis),

sunfish (warmouth, Lepomis gulosus; bluegill, L. macrochirus; redear

sunfish, L. microlophus), and shad (gizzard shad, Dorosoma cepedianum;

threadfin shad, D. petenense). Additional fish species, such as the

bottom-dwelling brown bullhead (Ictalurus nebulosus) and Florida gar

(Lepisosteus platostomus) also were captured during electrofishing

bouts, but were excluded from analysis. These species represent prey

functionally unavailable for ospreys, and I felt their inclusion in the

analysis would bias results by inflating preference measures of other

prey.

Prey species were categorized for several reasons. First, fish

species placed together are behaviorally and ecologically similar. For

instance, all three sunfish interbreed (Breder and Rosen 1966:439-440,










50

Childers 1967), as do both species of shad (Minckley and Krumholz 1960).

Second, foraging habits of the categorized species are similar

(references in Breder and Rosen 1966, Carlander 1969, 1977). Last, body

forms of grouped fish species are similar, and it is unlikely that

ospreys are capable of distinguishing minute differences (e.g., presence

of red dot on lateral surface of redear sunfish) between species in the

absence of gross behavioral differences. Thus, categories represent

groupings based primarily on prey ecology, not taxonomic considerations.

Prey relative abundances varied throughout both years (Fig. 3-1).

Sunfish were the predominant species in both seasons, comprising 48-65%

and 39-56% of the available prey in 1985 and 1986, respectively. The

most common size class was 20-30 cm. Sunfish 30-40 cm in length were

relatively uncommon. Shad had the second highest relative abundances in

both years. Shad in the 10-20 cm size class increased throughout the

post-fledging period, while 20-30 cm sized shad tended to decrease in

relative abundance. This most likely reflects the movement of young age

class fish from shallow edge waters into the lake at large. Bass were

the least common species and exhibited little yearly variation. In

general, patterns of change in relative abundance for all three prey

categories were similar for both years.

Absolute abundance of fish, measured by the number of fish captured

per 15-min electrofishing transect, exhibited patterns of change

slightly different from those of change in relative abundance (Fig.

3-2). Bass, for instance, exhibited a gradual decrease in absolute

abundance but remained fairly constant in relative abundance in 1985.

Most differences between change in absolute and relative abundance













Figure 3-1. Mean relative abundance per 15-min sample transect (n = 12) of each prey type and
size class during post-fledging periods in 1985 and 1986.
















1985

A--A 10-20 cm
20 --m 20-30 cm
S**-e 30-40 cm


20 -


40




" 20
)


........... ........... .......... ....... .. ....


i-~ -









. ..... .. *. .... ... ..

..0 60 90 120 10**


30 60 90 120 150


1986





-...........
i I I


--A^ ---m----

-












AA A--

S.. .......- ...........


30 60 90 120 150


DAYS FROM MEAN FLEDGING DATE













Figure 3-2. Mean number of each prey type and size class captured per 15-min sample transect
(n = 12) in 1985 and 1986.











1985
A--A 10-20 cm
10 I "-U 20-30 cm
A***** 30-40 cm

.... .:_.-- -.... .... ..-*


15 -


5



25 -


.. **... .. ............. ...


30 60 90 120 150


1986





. ...... ..... .........V


I I I I


. . . .... ....






. . .. ....... .......... .. ....


30 60 90


120 150


DAYS FROM MEAN FLEDGING DATE


i I m I I

-U o



-U -----5--
S.. .....
I m m I I


Z"A
Um ,


AA U~-.--


- ---- %~
A


'-*.^









55

occurred in sunfish in 1985. In May, the relative abundance of sunfish

20-30 cm in length was roughly twice that of the 10-20 cm size class.

Absolute abundance, in contrast, was fairly equal, with the 10-20 cm

size class only slightly more abundant. Furthermore, during the period

May-July when relative abundance of the 10-20 cm size class was

constant, absolute abundance steadily decreased. Thus, fledgling

ospreys were confronted by an environment whose potential prey resource

varied simultaneously in both relative and absolute abundance.

Yearly differences in absolute abundance were more pronounced.

With the exception of shad in the 10-20 cm size class, all prey and size

class combinations tended to decrease in 1985 and increase in 1986.



Fledgling Foraging Patterns

A total of 11 young in 1985 and 19 young in 1986 were color banded.

These birds represented 43% and 37%, respectively, of the yearly

reproductive output of the population on Newnan's Lake (Edwards,

unpublished data). Due to mortality factors and small sample sizes for

some fledglings, data for only 8 young in 1985 and 14 young in 1986 were

analyzed (Table 3-1).

Hereafter, short-hand notation will be used to describe the

relationship among the factors, where P represents the prey species

groups, Sthe prey size class, T days from the mean fledging date, and B

the individual bird. For example, the term PB states that prey choice

is associated with individual birds (i.e., the factors are not

independent of one another).












Table 3-1. Summary of capture data for 1985 and 1986. Capture data for
size class cell are collapsed across individuals.


each time period, fish species and


Bass Shad Sunfish

Size Class (cm): 10-20 20-30 30-40 10-20 20-30 30-40 10-20 20-30 30-40 Totals


1985 (8 birds)
Time (days)


10 2 0 11 4 0 35
15 14 1 18 19 3 35
22 17 1 12 15 2 12
16 15 0 20 14 1 21
16 14 2 25 21 1 18

79 62 4 86 73 7 121


96
155
115
104
120

590


1986 (14 birds)
Time (days)


77 50 3


123 342 3 1002


Totals


Totals









57

A variety of models adequately describes the ontogeny of foraging

in this population of ospreys (Table 3-2). There exists, however,

considerable difference in the adequacy of certain models between years.

For example, of the 15 statistically sound models, only PST and PST,PB

fit the data in both years. Neither model, however, was the best-fit

model in either year. Based on conditional tests (see methods), the

best-fit model also differed between years. The remaining models were

specific for each year.

Model PS,PB,PT. The best-fit model in 1985 was PS,PB,PT. This

model states that choice of prey by fledglings was related to prey size

class, individual bird, and time. In contrast, an entirely different

model best explained foraging in 1986. This model, PB,ST, states that

prey choice was related to individual birds, and that choice of size

class was related to time.

Use of the available size classes by fledglings in 1985 was

non-random, with young essentially ignoring fish 30-40 cm in length

(lambda = -1.014, SE = 0.101, P < 0.001). Use of the remaining two size

classes was at a rate greater than that expected at random (10-20 cm,

lambda = 0.517, SE = 0.064, P < 0.001; 20-30 cm, lambda = 0.497,

SE = 0.066, P < 0.001). In contrast, all three prey groups were

captured at random.

Examination of the PS interaction term reveals that difference in

use of size class was associated with sunfish (Table 3-3). Young tended

to use sunfish 20-30 cm in length at rates greater than expected at

random and ignored the larger-sized sunfish. The remaining prey groups

and size classes did not differ from random. Young apparently used











Table 3-2. Summary of models fitting foraging patterns of recently fledged ospreys in 1985 and 1986.
Missing values indicate a particular model did not adequately fit the data for that year.
Conditional tests were used to determine the best-fit model for each year. Best-fit models for
1985 and 1986 were PS,PB,PT and PB,ST, respectively. See text for explanation of model terms.


Model Description Year
85 86

Choice of prey is related to: G df G df


ST
ST,P
PB
PB,PS
PB,S
PB,ST
PS,PT
PS,BT
PS,PB,PT
PS,PT,BT
PS,PB,PT,BT
PST
PST,PB
PST,BT
PST,PB,BT


choice of size class related to time.
choice of size class related to time; varies over prey.
bird.
bird; size class.
bird; varies over size class.
bird; choice of size class related to time.
size class; time.
size class; choice by bird related to time.
size class; bird; time.
size class; time; choice by bird related to time.
size class; bird; time; choice by bird related to time.
both size class and time.
both size class and time; bird.
both size class and time; choice by bird related to time.
both size class and time; bird; choice by bird related to time.


245 271


223 280
234 257
169 259
177 249
136 235
183 259
129 238
137 228
97 214


522
520
495 U

493
481





495
456










59

Table 3-3. Lambda estimates and SE's for the best-fit model PS,PB,PT in
1985. Standardized lambda estimates can be obtained by
dividing each lambda estimate by its SE. Positive and
negative signs show direction of preference; significant
estimates signify use at a rate greater or less than that
expected at random. Superscripted birds with the same
number indicate siblings.


Prey-Size Interaction Term (PS)


Shad


Sunfish


lambda SE lambda


0.007
0.111
-0.118


0.096
0.140
0.096


-0.037
0.195
-0.158


SE lambda SE


0.088
0.131
0.088


0.030,
-0.306**
0.276


Prey-Bird Interaction Term (PB)


Shad


lambda SE lambda


-0.191
-0.264
-0.219
0.220
0.251
-0.255**
0.389
0.069


0.173
0.161
0.161
0.130
0.141
0.188
0.130
0.175


SE lambda SE


0.322 0.148
-0.242 0.152
-0.155 0.151
0.017 0.121
0.012, 0.146
-0.320 0.162
-0.164 0.138
-0.110 0.168


-0.131,**
0.506**
0.374
-0.237
-0.263
-0.065
-0.225
0.041


0.147
0.125
0.128
0.126
0.138
0.153
0.130
0.152


Prey-Time Interaction Term (PT)


Shad


Sunfish


lambda SE lambda SE lambda SE


-0.287 0.143
-0.125, 0.111
0.260 0.111
0.111 0.120
0.041 0.116


-0.209
-0.072
-0.151
0.171,
0.261


0.132
0.106
0.115
0.117
0.107


0.497, 0.109
0.197 0.093
-0.109, 0.106
-0.283*, 0.112
-0.301 0.105


*P<0.05; **P<0.01; ***P<0.001.


10-20
20-30
30-40


0.092
0.154
0.093


Sunfish


Bird

J01-851
J02-851
J03-852
J04-852
J05-852
J06-85
J07-85
J08-85


Time

30
60
90
120
150










60

these prey and size class combinations in relation to their

availability.

The presence of a PB interaction term suggests that considerable

variation exists among individual birds and their use of the available

prey resources. Four of the eight banded young in 1985 captured fish in

relation to their availability, while two concentrated on sunfish. The

remaining two birds concentrated on shad and bass, respectively. An

unexpected pattern was the lack of difference in preference between

siblings, even though there existed differences in how the sib-groups

used prey. Siblings J02-85 and J03-85 did not differ statistically in

their use of bass (z = -0.72, P > 0.05), shad (z = -1.69, P > 0.05) or

sunfish (z = -1.80, P > 0.05), although both young exhibited preference

for sunfish (Table 3-3). Similarly, siblings J04-85 and J05-85 had

identical preference patterns (bass, z = 0.28; shad, z = 0.49; sunfish,

z = 1.03; P > 0.05 for all three prey types). They did not, however,

exhibit preference for any prey type and instead used prey in relation

to their abundance.

The relationship between prey and time since fledging (PT) shows a

strong pattern of change, with young initially ignoring bass and

concentrating on sunfish. By 150 days into the post-fledging period,

however, young had changed emphasis and were exhibiting preference

towards shad.

Model PB,ST. Although a different model, PB,ST, met the best-fit

criteria for 1986, there exist some similarities in model terms between

years. Young in 1986 also ignored fish 30-40 cm in length

(lambda = -1.185, SE = 0.077, P < 0.001) and exhibited strong preference










61

for both the 10-20 cm (lambda = 0.509, SE = 0.05, P < 0.001) and 20-30

cm size classes (lambda = 0.676, SE = 0.048, P < 0.001).

The presence of a PB interaction suggests, as in 1985, that there

existed considerable variation in preference among young (Table 3-4).

Ten of 14 young used the available prey at random, although examination

of lambda estimate signs indicates considerable variability among young

in the direction of preference. Three birds, J06-86, J09-86, and

J13-86, showed strong preference for shad, taking this prey at a rate

greater than expected at random. Only J04-86 exhibited preference for

bass. In contrast to 1985, where preference for prey was divided

roughly among young, the majority of young in 1986 used available prey

in relation to their availability.

Siblings again exhibited similar preference patterns. Young J04-86

and J05-86 did not differ in their use of bass (z = -1.28, P > 0.05),

shad (z = -1.10, P > 0.05), and sunfish (z = 0.22, P > 0.05). Siblings

from one nest, however, did differ slightly. Fledgling J04-86 exhibited

strong preference for bass while its sibling, J05-86, did not.

Comparison of the lambda estimates, however, revealed that the

difference in use of bass between the two siblings was not significant

(z = 0.88, P > 0.05). Preference for shad (z = -0.94, P > 0.05) and

sunfish (z = 0.36, P > 0.05) also did not differ.

Unlike the best-fit model for 1985, there was no PT interaction

term in 1986. Instead, preference for particular prey types remained

constant throughout the entire post-fledging period. Both shad

(lambda = 0.188, SE = 0.044, P < 0.001) and sunfish (lambda = 0.4,












Table 3-4. Lambda estimates and SE's for the best-fit model PB,ST in
1986. Standardized lambda estimates can be obtained by
dividing each lambda estimate by its SE. Positive and
negative signs show direction of preference; significant
values signify use at a rate greater or less than that
expected at random. Superscripted birds with the same
number are siblings.


Prey-Bird Interaction Term (PB)


Shad


Bird

J01-86
J02-86
J03-861
J04-861
J05-861
J06-862
J07-862
J08-862
J09-86
J10-86
J11-86
J12-86
J13-86
J14-86


Sunfish


lambda SE lambda SE lambda SE


-0.033
0.003
-0.082,
0.348
0.288
-0.076
0.072
0.164
-0.250
0.270
-0.066
-0.151,
-0.469
-0.018


0.166
0.150
0.236
0.167
0.155
0.179
0.181
0.204
0.194
0.165
0.175
0.188
0.229
0.254


Size-Time Interaction (ST)


10-20 cm


-0.028
-0.119
-0.022
-0.269
-0.188,
0.300,
-0.282,
-0.373,
0.280
-0.165
0.166
0.259,,,
0.601
-0.160


0.131
0.138
0.183
0.156
0.140
0.134
0.141
0.188
0.140
0.147
0.134
0.140
0.151
0.207


0.061
0.116
0.104
-0.079
-0.100
-0.224
0.211
0.209
-0.030
-0.105
-0.100
-0.108
-0.132
0.178


0.126
0.128
0.176
0.143
0.131
0.142
0.138
0.160
0.140
0.140
0.134
0.144
0.162
0.190


Size


20-30 cm


30-40 cm


lambda SE lambda SE lambda SE


-0.164
0.105
0.061
-0.045
0.043


0.095
0.099
0.096
0.098
0.090


0.361 0.146
-0.267 0.166
-0.121 0.158
0.034 0.159
-0.007 0.135


*P.; **P0.; 0.0.
P<0.05; P<0.01; P

Time

30
60
90
120
150


-0.197
0.162
0.060
0.011
-0.036


0.100
0.102
0.099
0.100
0.093










63

SE = 0.044, P < 0.001) were preferred prey types, while bass were

ignored (lambda = -0.588, SE = 0.054, P < 0.001).

Patterns derived from the ST interaction are less clear.

Difference in use of the available prey size classes appears restricted

to the first 30 days of post-fledging, when young exhibited strong

preference for the smallest size class. Fish in the 30-40 cm size class

were ignored.



Fledgling Versus Adult Foraging Patterns

Comparison of use of available prey by young with adults revealed

differences in learning curves between years, even though young of both

years eventually exhibited patterns of prey preference similar to that

of adults. Implicit in these results is the assumption that adults

represent some baseline of proficiency that is better than that

exhibited by naive fledglings.

Examination of change in preference of young in 1985 revealed a

gradual shift in choice, until by 150 days young were foraging in a

manner similar to adults (Fig. 3-3). There exist, however, differences

in the length of time required to reach the level of preference

exhibited by adults for each prey class. Young fell within confidence

limits of adult preference for bass by 90 days, while it took 150 days

to attain preference levels exhibited for shad and sunfish. Use of prey

size classes also differed between adults and fledglings in 1985 (Table

3-5). In general, young captured smaller size class prey than did

adults.













Figure 3-3. Change in preference of available prey types by fledglings (solid squares) in
1985 in relation to time. Rectangles represent 95% confidence intervals about
the lambda estimates of adult data fitted to the model PS,PB,PT. Asterisks
represent significant difference in preference patterns between young and adults.













Bass


30




60


Shad


"-------^


90


0 120

C)
>-\
150



-1.0 0 1.0


-1.0 0 1.0


LAMBDA ESTIMATE


Sunfish



























-1.0 0 1.0










66

Table 3-5. Lambda estimates and SE's for adult data fitted to the
best-fit model PS,PB,PT in 1985. Positive and negative
signs within parentheses indicate significant difference in
use of particular prey and size class combinations between
young and adults. Signs also indicate direction of
preference.


Adult Prey-Size Interaction Term (PS)

Bass Shad Sunfish

lambda SE lambda SE lambda SE
Size

10-20 cm 0.207 0.145 -0.260 (+) 0.108 0.054 0.134
20-30 cm 0.583 (-) 0.233 0.129 0.181 -0.713 0.241
30-40 cm -0.790 (+) 0.181 0.131 (-) 0.121 0.659 (-) 0.144










67

The lack of a PT interaction in the best-fit model for 1986 implies

that no relationship existed between time since fledging and prey choice

in young. Young exhibited the same pattern of prey preference from

fledging until they dispersed from the area, and did not differ from

adults in use of prey (bass, z = -1.44, P > 0.0; shad, z = 1.88,

P > 0.05; shad, z = 1.453, P > 0.05) at any time during the

post-fledging period. In contrast to 1985, prey preference patterns of

young in 1986 did not differ from that exhibited by adults during the

same time period.



Discussion



The lack of similarity in best-fit models between years suggests

that no one pattern best describes the ontogeny of foraging in this

population of ospreys. This difference may be due to between year

differences in prey absolute and relative abundance.

Young in 1985 exhibited a strong temporal shift in preference,

changing from preference for sunfish soon after fledging to shad by the

end of the post-fledging period. Initially, sunfish were the most

abundant prey type. By 150 days, however, the absolute abundance of

sunfish had decreased, and young were concentrating on shad. The switch

to shad parallels an increase in their absolute abundance. Choice by

young ospreys reflects these changes in absolute abundance, suggesting

that young may be more sensitive to absolute than to relative abundance

of prey.










68

Sensitivity to change in prey absolute abundance is a factor common

to foraging models concerned with energy optimization (MacArthur and

Pianka 1966, Schoener 1971, Pulliam 1974, Werner and Hall 1974, Charnov

1976a). These models predict that an increase in absolute abundance of

preferred prey results in an increased representation of those prey

types in the diet. While it is tempting to conclude that young ospreys

foraged in a manner consistent with energy-optimizing models, several

factors make such a conclusion tenuous at best.

A major complication is that my results also provide general

support for prey switching models. These models (Murdoch 1969, 1973,

Murdoch and Oaten 1975) predict a shift in preference as relative

abundance changes, with predators concentrating on the most abundant

prey type. Because shifts in relative abundance in 1985 roughly

parallel shifts in absolute abundance (Figs. 2-1 and 2-2), it is

difficult to ascertain whether young are responding to change in

absolute (energy-optimizing) or relative (nutrient-optimizing)

abundance.

Second, it frequently is difficult to determine the preferred or

"higher-ranked" prey from an energetic sense. Here, for instance, I was

unable to determine if costs (e.g., handling) and benefits differed

among the available prey types. While it is easy to define preference

based on observations of choice by young ospreys, it is difficult to

determine a priori whether sunfish represent "higher-ranked" prey.

Last, it is unlikely that naive young know what constitutes

"higher-ranked" prey when they fledge. Such knowledge requires a fairly

rigid genetic basis that is not likely in a species confronting a










69

variable environment (Glasser 1982, 1984; Glasser and Price 1982). It

is more likely that young ospreys are simply exhibiting preference for

the most common prey type, initially selecting for sunfish due to their

high absolute and relative abundance. Similarly, the shift in

preference to shad was associated with a concurrent increase in its

absolute and relative abundance. Both shifts in preference indicate a

high degree of flexibility in preference patterns of young ospreys.

Evidence for flexibility in foraging within this population of

ospreys is strengthened by results from 1986. The lack of a PT

interaction in this year implies choice of prey by young was constant

over the course of post-fledging. Thus, although absolute and relative

abundance of prey types and size classes varied over the post-fledging

period, choice by young remained constant. Constant preference for prey

types irrespective of variation in prey abundance is an explicit

prediction of foraging models based on nutrient constraints (Marten

1973, Pulliam 1975, Westoby 1978). That results from different years

tend to support different groups of foraging models is not surprising,

and lends credence to the concept that prey choice should vary with

fluctuations in the prey resource base. Predators, and in particular

young in the process of developing foraging skills, should not be

expected to operate within the constraints of particular foraging

models. Young birds, for instance, require a certain degree of

flexibility so they can react to whatever environmental conditions they

initially face. Such flexibility, particularly as it relates to the

ability to respond to new or highly variable environments, is clearly










70

advantageous to a species whose young must soon migrate from the natal

area.

A common factor to the best-fit models of both years was the PB

interaction term. This term suggests individuals differed in use of the

available prey types (see also Murdoch and Oaten 1975, Chesson 1984).

The presence of a PT interaction term in 1985 also suggests this

individual variation in preference was related to time. The lack of a

PT interaction in 1986, however, suggests that individual young did not

change preference over time. Instead, differences among young were

constant over time. Whether these differences among young, with some

individuals foraging at random and others exhibiting strong preference

for one or more prey types, affect subsequent survivorship is not known.

Clearly, flexibility in foraging exists both at a population level

between years, and among young within each year.

Learning curves of young also differed between years. Young in

1985 did not attain preference patterns of adults until at least 120

days into the post-fledging period. Learning curves in this year showed

shifts in preference over time. In contrast, young in 1986 almost

immediately "locked into" the foraging patterns of adults, and did not

vary preference patterns throughout post-fledging. A contributing

factor to this difference in learning curves may be between year

differences in prey abundance.

Prey abundance in 1986 was increasing when young fledged. An

increase in the number of available prey during the period young are

developing foraging skills may increase encounter rates, thereby

allowing for finer discrimination among prey types. The ability to










71

discriminate may develop if young are able to retain whatever cues led

to a successful foraging bout. Retention of successful cues may

represent a variation of the "win-stay, lose-shift" response strategy

common in the learning set literature (e.g., Levine 1959). Although its

ecological application has been primarily to predator patch choice

(e.g., Cole et al. 1982, Wunderle and O'Brien 1985), the principle of

repeating when successful or avoiding when unsuccessful is applicable to

the development of preference for prey in naive young. A constant or

increasing prey resource provides young with the opportunity to rapidly

develop discrimination for prey types, even if foraging skills, as

measured by indices such as capture success, are low.

In contrast to 1986, decreasing prey abundance in 1985 may have

forced young ospreys to abandon regularly whatever cues developed from a

"win-stay, lose-shift" strategy. Rewards were not sufficiently

consistent to warrant cue retention. Even if successful cues were

retained, lower encounter rates may have made it difficult for young to

exhibit constant preference for particular prey types. Lower encounter

rates coupled with lower success rates may have forced young to begin

sampling again instead of waiting for cues associated with "win-stay."

Subsequent successful capture of a different prey type would lead to the

retention of an entirely new set of cues. Thus, change in preference

over time may represent shifts in learning sets as young are constantly

forced to acquire new cues.














CHAPTER IV


SIBLING ENHANCED FORAGING IN OSPREYS



Introduction



Differences in the foraging ability of adult and young birds have

been documented for a variety of avian species (references in Brandt

1984). One explanation may be that the coordination and motor skills of

young are not fully developed. Consequently, young may be less

effective than adults in the mechanics of prey capture (e.g., Ingolfsson

and Estrella 1978, Brandt 1984). Alternatively, young may be as

proficient as adults in the mechanics of foraging, but lack knowledge of

available resources (e.g., Davies and Green 1976). Young simply have

not had sufficient opportunity to encounter prey items and assimilate

cues leading to successful foraging behaviors.

Here, I examine the role of social relationships in shaping the

development of foraging skills in fledgling ospreys (Pandion haliaetus).

An important component of this process may be potential benefits derived

from interactions with other young and adult ospreys. One benefit of

sociality may be an increased rate of learning due to the more frequent

exposure to successful foraging cues. In contrast, young learning to

forage without the opportunity to interact with conspecifics may develop

foraging skills at a slower rate.

72










73

I also examine the nature of interactions among young and between

adult and young ospreys during the post-fledging period. One factor

thought to influence foraging performance is competition (MacArthur

1972, Werner 1976), and young ospreys might be expected to exhibit

considerable aggression towards other young or adults if resource

availability is low or if they initially have a difficult time capturing

prey. If, however, there exist benefits related to observational

learning, then the frequency of aggressive interactions between young

might be lower.



Study Area and Methods



My research was conducted on Newnan' Lake, Alachua County, Florida,

during April-September 1985 and 1986. Newnan's Lake, located 15 km east

of Gainesville, Florida, is a 2400 ha hyper-eutrophic lake (Shannon and

Brezonik 1972) rimmed with baldcypress (Taxodium distichum) and mixed

hardwoods. Most of Newnan's Lake is open water, although from May to

October much of the lake periphery is covered by coontail (Ceratophyllum

spp.) and spatterdock (Nuphar luteum) vegetation.

To identify young, I climbed suitable nest trees and marked

nestling ospreys with unique color band combinations prior to fledging.

I collected data on fledgling ospreys from a boat anchored offshore at

locations that facilitated simultaneous observations on several birds.

Individual fledglings were chosen at random and watched until the

completion of a 15-min period or until a successful capture was made. I

preferred this approach over continuous observations due to difficulties










74

associated with maintaining observation of the same individual, and

because it increased the number of independent samples. Instances where

I was unable to identify individuals could were excluded from analysis.

Observations were collected during two 10-day periods each month.

Behavioral observations were standardized to days since fledging for

each bird and placed in 30, 60, 90, 120, and 150-day categories. I

considered young to have fledged once they left the nest tree.

Hereafter, "related" refers to two young from the same nest and

"sibling-group" to related birds engaged in social or hunting activities

together. "Unrelated" birds were those from single chick nests. "Solo"

refers to a hunting flight by a single bird; "tandem-hunt" to hunts

where more than one bird participated in a hunting flight. Both related

and unrelated birds participated in solo and tandem-hunt flights.

Criteria defining participation included close (<10 m) proximity,

similar flight direction and foraging height, and capture attempts by

all birds directed towards the same prey item. Observation of a

tandem-hunt was terminated if one of the birds broke off hunting and

left the area.

A capture attempt included any stooping effort that brought the

bird within 3 m of the water surface. I defined successful capture as

the capture and retention of a prey item for at least 15 s or until the

bird landed in a tree to begin feeding. Unsuccessful capture attempts

included any attempt at capture that failed to result in retention of a

prey item. Loss of prey items within the first 15 s was due primarily

to pirating attempts by other birds or the inability of the bird to










75

grasp and control the fish. I excluded losses of this nature from the

calculation of individual capture success rates.

Multiple attempts for the same prey item each constituted one

capture attempt. I evaluated capture attempts by sibling-groups in two

ways. First, capture success was evaluated for the sibling-group. As

before, each attempt by a bird during a tandem-hunt was counted. I also

calculated the capture success of each individual involved in a

tandem-hunt.

Data were analyzed using weighted G-tests (Sokal and Rohlf 1981).

The level of significance for all analyses was 0.05.


Results



A total of 11 young in 1985 and 19 young in 1986 were color banded.

Due to mortality factors and small sample sizes for some fledglings,

data for only 8 young in 1985 and 14 young in 1986 were analyzed.

Fledgling ospreys attempted to capture prey as early as five days after

fledging, but the earliest successful capture I observed was 11 days

after fledging. All banded birds had made at least one successful

capture by 20 days into the post-fledging period.

Capture success for each 30-day category did not differ between

years for either related or unrelated young. Consequently, data for

both years were combined for analysis. Capture success of ("related")

sibling-groups and ("unrelated") lone ospreys was initially similar, but

by 60 days from the mean fledging date sibling-groups were more

successful at capturing prey than were lone ospreys (Table 4-1).











Table 4-1. Comparison of capture success by related and unrelated
fledgling ospreys in relation to days from mean fledging
date (DMFD). Data for 1985 and 1986 did not differ for
each 30-day category and were combined for analysis.
A = Number capture attempts observed; S = Number
successful captures.


DMFD Related Unrelated z-score

A S (%) A S (%)


30 333 73 (21.9) 742 162 (21.8) 0.03

60 388 132 (34.0) 838 234 (27.9) 2.17

90 181 104 (57.4) 687 283 (41.2) 3.90

120 194 133 (68.6) 351 181 (51.6) 3.84

150 214 143 (66.8) 313 183 (58.5) 1.92

P < 0.05; P < 0.001










77

Differences were greatest at 90 days post-fledging. Capture success of

both groups was statistically similar by 150 days, although capture

success of lone birds was still lower than that of sibling-groups.

Differences also existed in the time required until capture success

of young was similar to that of adults. Capture success by

sibling-groups approached that of adults at a faster rate than that

exhibited by unrelated birds foraging by themselves (Fig. 4-1).

Capture success also varied with type of hunting (Table 4-2). More

than 80% of the capture attempts by related birds occurred during tandem

hunts. Birds with siblings seldom hunted by themselves. In contrast,

even though unrelated birds occasionally hunted and attempted to capture

prey together, more than 90% of capture attempts by lone birds occurred

during solo hunts. Unrelated birds seldom remained in close proximity

to one another, particularly during hunting flights. Capture success by

unrelated birds hunting in tandem was consistently lower than that of

unrelated birds hunting singly (Table 4-2), although differences were

not significant (P > 0.05 for each 30-day category). Capture success by

related birds was similar regardless whether they hunted in tandem or

singly.

Young tended to remain near and interact with adults during the

first 30-60 days post-fledging (Fig. 4-2). As the post-fledging period

progressed, young interacted less with adults and more with other young.

The number of observed interactions dropped considerably just prior to

the time young dispersed from the area.

The nature of interactions between adults and young and between

young varied over the post-fledging period. In general, interactions













Figure 4-1. Mean capture success of related and unrelated ospreys compared to 95% confidence intervals
about adult capture success during the same time period. Data for 1985 and 1986 are
combined.















S80 -







CO
w
Ix 40 -
S 4o 0 Adult 95% CIl
0 o Related
O0 0 Unrelated
20 -



30 60 90 120 150


DAYS FROM MEAN FLEDGING DATE










Table 4-2. Comparison of tandem and solo hunt capture success of related and unrelated fledgling
ospreys in relation to days from mean fledging date (DMFD). Data for 1985 and 1986 are
combined. A = Number capture attempts observed; S = Number successful captures.


Tandem Hunts Solo Hunts
Related Unrelated Related Unrelated
DMFD A S (%) A S (%) z-score A S (%) A S (%) z-score

30 297 64 (21.5) 27 4 (14.8) 0.92 36 9 (25.0) 715 158 (22.1) 0.39
60 332 114 (34.3) 41 12 (29.3) 0.66 56 18 (32.1) 797 222 (27.8) 0.67
90 159 93 (58.4) 58 19 (32.8) 3.51 22 11 (50.0) 629 264 (41.9) 0.75
120 155 106 (68.4) 42 19 (45.2) 2.71 39 27 (69.2) 309 162 (52.4) 2.12
150 126 85 (67.5) 37 19 (51.4) 1.75 88 58 (65.9) 276 164 (59.4) 1.11

P < 0.05; P < 0.01; P < 0.001













Figure 4-2. Number of observed fledgling-fledgling and adult-fledgling interactions in relation
to days from mean fledging date. Number of interactions is weighted by sample
effort. Data for 1985 and 1986 are combined.













W Adult -
Fledgling
Fledgling-
Fledgling


I 1 t


60


DAYS FROM MEAN FLEDGING DATE


10




5


7
7

77






'K


7
7//


F-









83

between young were more aggressive than affiliative in nature.

Aggressive actions directed towards other young included pirating

attempts, perch displacements, and beak and foot stabs. Of interest,

however, is the large difference in the proportion of aggressive

interactions initiated between related and unrelated young (Table 4-3).

Related birds had fewer aggressive interactions with other young than

did unrelated birds, even though related birds were in the proximity of

one another more often than related birds met. Roughly 90% (50 of 55)

of observed aggressive interactions between related young involved

fights over a prey item captured by, or in the possession of, one of the

young.

In contrast, related and unrelated birds did not differ greatly in

the manner in which they interacted with adults (Table 4-3).

Interactions with adults were more aggressive than interactions with

other fledglings. Unlike fledgling-fledgling interactions, however,

aggressive interactions with adults primarily involved attempts to

pirate prey. Young rarely attempted to displace adults from perches or

initiate other kinds of aggressive interactions.


Discussion



There exist several processes by which naive young incorporate

information from initial foraging bouts into a successful foraging

strategy. One method is trial and error, often referred to as a

"win-stay, lose-shift" response strategy in some learning set literature

(Levine 1959, Kamil and Yoerg 1982). Through repeated sampling of their









84
Table 4-3. Frequency of aggressive interactions between fledglings and
between fledgling and adult ospreys in relation to days
from mean fledging date (OMFD). Fledglings are categorized
as either related or unrelated (see text). Data for 1985
and 1986 are combined. T = Total number of affiliative and
aggressive interactions observed; N = Number of aggressive
interactions observed.


Unrelated Related
DMFD T N (%) T N (%) z-score

Fledgling-Fledgling Interactions
30 13 8 (61.5) 10 3 (30.0) 1.59
60 23 17 (73.9) 15 5 (33.3) 2.66
90 89 45 (50.6) 55 14 (25.4) 3.19
120 125 59 (47.2) 67 23 (34.3) 1.76
150 32 14 (43.8) 29 10 (34.5) 0.75

Adult-Fledgling Interactions
30 91 72 (79.1) 47 34 (72.3) 0.87

60 96 78 (81.2) 69 61 (88.4) -1.29
90 53 33 (62.3) 34 28 (82.3) -2.14
120 25 13 (52.0) 17 7 (41.2) 0.69
150 26 7 (26.9) 11 4 (36.4) -0.56

P < 0.05; P < 0.01










85

environment, young eventually learn to recognize and retain cues

associated with successful foraging bouts. Cues associated with

unsuccessful foraging bouts are not retained. Another method increasing

foraging ability is observational learning (Turner 1964), whereby naive

birds use conspecifics as role models. Here, birds not only learn

through individual trial and error, but also by simultaneously observing

and retaining those cues leading to successful foraging bouts by other

birds. Observational learning, when possible, is considered a more

efficient method than trial and error for developing the skills

necessary to exploit novel resources (Galef 1976).

The importance of observational learning in the development of

foraging behavior in fledgling ospreys can only be inferred from the

data presented here. That it plays some role, however, is supported by

several lines of evidence. First, even though considerable variation in

how fledglings used available resources existed, siblings, who tended to

remain together throughout the post-fledging period, had similar

patterns of prey preference (Chapter III). In addition, related birds

hunting together not only had greater success at capturing prey, but

their success rate also approached that of adults at a faster rate than

that exhibited by unrelated birds. The more rapid development of

foraging skills in naive young having opportunity for observational

learning has been documented for a variety of bird (Turner 1964, Dawson

and Foss 1965, Alcock 1969) and mammal species (references in Weigl and

Hanson 1980). In general, role models in these studies were experienced

animals presumably foraging at some optimal level. Naive animals

apparently learned appropriate foraging techniques through observation.










86

In contrast, it appears that naive ospreys are capable of learning

appropriate foraging behaviors from other inexperienced birds as well as

from experienced birds. The rapid increase in capture success by

related birds foraging together suggests that exposure to mistakes may

be as beneficial as exposure to successful capture attempts, even when

the individual bird is not directly involved in the capture attempt.

Rapid development of foraging behaviors by young ospreys may be

important for several reasons. Although the majority of adults at

Newnan's Lake are year-around residents, young of each year disperse

from the area by roughly 150 days post-fledging and presumably migrate

to Central America. Birds having properly developed foraging skills may

have increased survivorship during migration relative to those that do

not, particularly since migrating birds continuously face new and

different foraging environments.

Different foraging environments also require rapid recognition of

potential prey items. Davies and Green (1976) showed that for the Reed

Warbler (Acrocephalus scirpaceus) the main factor limiting foraging

performance was not mechanics, but rather recognition that flies

constituted food items. Migrating ospreys undoubtedly face similar prey

recognition problems, and may require a period of adjustment when they

encounter a new foraging environment. Observations that related young

learn to recognize profitable prey sooner than unrelated young (Chapter

III) suggest that observational learning not only aids development of

foraging mechanics, but also prey identification. The rapidity with

which migrating birds learn to recognize profitable prey items could, as

before, affect survivorship, particularly when temporal constraints on










87

the migration period exist. Whether related young ospreys remain

together through all or any part of migration, however, is not known.

The decreased frequency of aggressive interactions between related

young also suggests there exists some level of cooperation between

related birds. The level of aggressive interactions between related

young remained fairly low and constant throughout post-fledging.

Aggressive interactions between unrelated young were high initially, but

decreased to a level similar for related young by 150 days

post-fledging. Decreased aggression between related young may be

necessary to realize benefits associated with social foraging (e.g.,

Pulliam 1973, Bertram 1978). Although predatory birds such as ospreys

rarely "flock" together, social foraging or "conspecific cueing "

(Kiester and Sklatkin 1974) by related birds may increase levels of prey

detection. Benefits related to predator avoidance are not applicable to

ospreys. Similar levels of aggression between related and unrelated

young once a prey item is captured, however, suggest that any

cooperation between related birds is transitory.
















CHAPTER V


CONCLUSION AND SYNTHESIS



While the concept of a facultative foraging strategy generally

describes the foraging behavior of this population of ospreys, my

observations do not entirely corroborate the concept as outlined by

Glasser (1982, 1984). They do, however, suggest it provides a more

general framework in which foraging behaviors can be evaluated.

Unfortunately, increased generality decreases the utility of the

concept as a rigorous approach to understanding foraging behaviors, a

not uncommon criticism of many foraging models. Moreover, the

predictions Glasser (1982, 1984) generates are not mutually exclusive

from other model predictions (e.g., Schoener 1971, Pulliam 1974).

My findings that ospreys change prey preference throughout the

year are not new. These observations, however, are enhanced by my

extensive documentation of the dynamics of the fish resource base. The

strong association between changes in the fish resource base and shifts

in prey preference suggests that adult ospreys at least are capable of

modifying foraging behavior in response to changing resource

conditions. Whether these seasonal shifts from a generalist strategy

in pelagic habitat to a specialist strategy in littoral habitat










89

constitute an "optimal" foraging strategy that maximizes fitness is

debatable.

Problems abound in current optimal foraging literature, ranging

from serious criticism that it is tautological (Ollason 1980) to

questions whether an optimal framework is even necessary to understand

foraging behavior (Heinrich 1983). That my results adequately fit a

variety of proposed models provides additional evidence that many

foraging models are too narrowly defined. To attempt to force my

results into a particular model adds little understanding to foraging

behavior, except to generate yet another set of ad hoc explanations for

deviations from model predictions. The ecology of osprey foraging

behavior may be understood best by using the heuristic value of optimal

foraging theory as a framework for interpretation of results.

In particular, my interpretation of the shift from one foraging

strategy to another was confounded by the presence of time lags. These

temporal lags in change of preference are not unlikely in predators

foraging in variable environments, and may be due to delay in the

recognition by the predator that previously ignored prey types now

constitute profitable food. This recognition may require the formation

of new search images that increase the predator's ability to recognize

and capture different prey types. Consequently, predators may continue

to capture less profitable prey types until a new search image is

formed. Similarly, the time required to develop a new search image

could explain why ospreys here did not immediately respond to increases

in the abundance of sunfish. Instead, change in preference was

gradual. Thus, rather than the abrupt shift in preference predicted by











some foraging models (e.g., Krebs et al. 1978), change in preference

could occur gradually. Lag in preference could simply represent time

required to develop a new search images appropriate for a different

resource base.

Habitat switches exhibited by my study population suggest an

opportunistic response by ospreys to increases in prey abundance.

Ospreys switched from foraging in pelagic to littoral habitat when prey

abundances in the littoral habitat increased and exceeded those in

pelagic habitat. The regularity associated with this pattern suggests

some form of frequency-dependent diet (Fullick and Greenwood 1979,

Horsley et al. 1979) best may describe osprey foraging. That temporal

lags do not disappear between years suggests that some form of memory

decay regarding preference occurs between years. Apparently, constant

reinforcement of a specific search image is necessary to maintain

adequate levels of prey detection (Pietrewicz and Kamil 1979, McNair

1981).

Temporal variation in the prey resource base also played an

important role in the ontogeny of foraging in this population of

ospreys. Differences in developmental patterns were strongly

associated with between-year differences in the prey resource base.

Decreasing prey abundance during the post-fledging period in 1985

extended the learning period by making it difficult for young to

develop appropriate learning sets. In contrast, increasing prey

abundance during 1986 shortened the learning period. Apparently, naive

ospreys require constant reinforcement of cues associated with











particular prey types before specific search images for prey can

develop.

There appears to be greater flexibility in diet choice than in the

mechanics of foraging. For instance, data related to the mechanics of

foraging (e.g., capture success) were more constant between years than

were estimates of prey preference. Such flexibility in the ontogeny of

diet choice is, of course, beneficial to naive predators entering

variable environments. In contrast, mechanical aspects of foraging

behavior should not be expected to exhibit such variation. Their

importance to survival dictates they rapidly be incorporated into a

predator's behavioral repertoire.

Social behavior also affected how ospreys developed foraging

skills. Faster rates of learning occurred when young had opportunity

to interact with other young. Because all young studied here

eventually attained foraging skills comparable to adults, benefits

associated with faster learning are hard to envision. Such may not be

the case for osprey populations farther north, where shorter summers

result in shorter post-fledging periods. There, considerable fitness

benefits may be associated with the faster development of foraging

skills. Young forced to leave their natal areas prior to the

development of a proper foraging strategy may be subject to increased

mortality risks, making rapid development of foraging mechanics of

vital importance. In contrast, longer summers in Florida allow for

longer post-fledging periods and time in which to develop a proper

foraging strategy. Whether benefits associated with faster learning









92

are important during post-fledging dispersal, however, is not known,

and poses an interesting question for future research.

Finally, the strong association between rates of learning and

social interactions has important management implications. Osprey

reintroduction programs are increasing in number and scope along the

Eastern Seaboard in an effort to return ospreys to areas they once

occupied. Reintroduction programs striving to maximize survival

probabilities of released birds may have greater success by releasing

young in clusters rather singly. At a minimum, complete broods of at

least two rather than single nestlings should be taken from donor

populations. Related young subsequently should be hacked from the same

tower. Separating related young eliminates potential benefits

associated with faster learning rates, benefits that may increase

survivorship of released young and increase the likelihood that ospreys

will once again be common.















LITERATURE CITED


Agresti, A. 1984. Analysis of ordinal categorical data. John Wiley
and Sons, New York, New York, USA.

Alcock, J. 1969. Observational learning in three species of birds.
Ibis 111:308-321.

Baker, M. C., and A. E. M. Baker. 1973. Niche relationships among six
species of shorebirds on their wintering and breeding ranges.
Ecological Monographs 43:193-212.

Bertram, B. C. R. 1978. Living in groups: predators and prey. Pages
64-96 in J. R. Krebs, and N. B. Davies, editors. Behavioural
ecology: an evolutionary approach. Sinauer Associates,
Incorporated, Sunderland, Massachusetts, USA.

Bishop, Y. M. M., S. E. Feinberg, and P. W. Holland. 1975. Discrete
multivariate analysis. Massachusetts Institute of Technology
Press, Cambridge, Massachusetts, USA.

Brandt, C. A. 1984. Age and hunting success in the brown pelican:
influences of skill and patch choice on foraging efficiency.
Oecologia (Berlin) 62:132-137.

Breder, C. M., Jr., and D. E. Rosen. 1966. Modes of reproduction in
fishes. The Natural History Press, Garden City, New York, USA.

Brown, M. B. 1976. Screening effects in multi-dimensional contingency
tables. Applied Statistics 25:37-46.

Buckley, F. G., and P. A. Buckley. 1974. Comparative feeding ecology
of wintering adult and juvenile royal terns (Aves: Laridae,
Sterninae). Ecology 55:1053-1063.

Caraco, T. 1980. On foraging time allocation in a stochastic
environment. Ecology 61:119-128.

Carlander, K. D. 1969. Handbook of freshwater fishery biology. Volume
1. The Iowa State University Press, Ames, Iowa, USA.

___. 1977. Handbook of freshwater fishery biology. Volume 2. The
Iowa State University Press, Ames, Iowa, USA.




University of Florida Home Page
© 2004 - 2010 University of Florida George A. Smathers Libraries.
All rights reserved.

Acceptable Use, Copyright, and Disclaimer Statement
Last updated October 10, 2010 - - mvs