-
Foraging patterns of kestrels and shrikes andtheir relation to
an optimal foraging model
Item Type text; Dissertation-Reproduction (electronic)
Authors Mills, Gregory Scott
Publisher The University of Arizona.
Rights Copyright © is held by the author. Digital access to this
materialis made possible by the University Libraries, University of
Arizona.Further transmission, reproduction or presentation (such
aspublic display or performance) of protected items is
prohibitedexcept with permission of the author.
Download date 05/04/2021 01:12:16
Link to Item http://hdl.handle.net/10150/565434
http://hdl.handle.net/10150/565434
-
© 1979
GREGORY SCOTT M I L L S
A L L RIGHTS RESERVED
-
FORAGING PATTERNS OF KESTRELS AND SHRIKES AND THEIR RELATION TO
AN OPTIMAL FORAGING MODEL
byGregory Scott Mills
A Dissertation Submitted to the Faculty of theDEPARTMENT OF
ECOLOGY AND EVOLUTIONARY BIOLOGY
In Partial Fulfillment of the Requirements For the Degree
ofDOCTOR OF PHILOSOPHY
In the Graduate CollegeTHE UNIVERSITY OF ARIZONA
19 7 9
Copyright 1979 Gregory Scott Mills
-
THE UNIVERSITY OF ARIZONA
GRADUATE COLLEGE
I hereby recommend that this dissertation prepared under my
direction by ___________ Gregory Scott
Mills______________________
entitled FORAGING PATTERNS OF KESTRELS AND SHRIKES AMD THEIR
RELATION TO AN OPTIMAL FORAGING MODEL__________________
be accepted as fulfilling the dissertation requirement for
the
degree of ______________ Doctor of
Philosophy_____________________
Dissertation Director Date
As members of the Final Examination Committee, we certify
that we have read this dissertation and agree that it may be
presented for final defense.
C }j-Iv ______ __f z /s ^ y 'T 'S
... u^u_ . ' v. . ' Lz-v ________ Jijzc* 7 g____l j . l U L . r
̂ ??
Final approval and acceptance of this dissertation is contingent
on the candidate's adequate performance and defense thereof at the
final oral examination.
-
STATEMENT BY AUTHOR
This dissertation has been submitted in partial fulfillment of
requirements for an advanced degree at The University of Arizona
and is deposited in the University Library to be made available to
borrowers under rules of the Library*
Brief quotations from this dissertation are allowable without
special permission, provided that accurate acknowledgment of source
is made* Requests for permission for extended quotation from or
reproduction of this manuscript in whole or in part may be granted
by the copyright holder*
SIGNED; 5
-
ACKNOWLEDGMENTS
I would like to thank Ho R=, Pulliam, Co R= Tracy, Jc Re Silli-
man, W0 Ao Calder, Stephen Mo Russell, and, especially, Jo Ho Brown
for their comments and suggestions concerning the ideas presented
in this papero Steve Sutherland made valuable contributions to the
concept of optimal perch height and Tom Caraco kindly shared his
ideas on risk aversiono Jo Ho Brown and Ao Co Gibson made valuable
contributions to the preparation of the manuscripto I thank the
staff and officers of The Research Ranch for their consent and aid
in some aspects of this study0
-
TABLE OF CONTENTS
PageLIST OF ILLUSTRATIONS o o o o o o o o o o o o o o o o o o
vLIST OF TABLES © © © © © © © © © © © © © © © © © © © © ©
vxABSTRACT © © o o © © © © © © © © © © © © © © © © © © © © ©
vxx
1© INTRODUCTION © © © © © © © © © © © © © © © © © © © © © © ©
I2© PATTERNS OF HUNTING FROM PERCHES © © © ............. © © 4
An Equation for Net Energy Gain © © © © © © © © © © © 4Metiiods
© o © © © © © © © © © © © © © © © © © © © © © ^Patch Choxce © o © ©
© © © © ©©© © © © © © © © © © © 6
Geometry of Hunting from Perches © © © © © © © © © 7Optxmal
Perch Hexght © © © © © © © © © © © © © © © 12Predictions and Tests
© © © © © © © © © © © © © © 19
Movement Between Patches © © © © © © © © © © © © © © ©
27Allocation of Time in Patches © © © © © © © © © © © © 30Opt xmal
Dxet © © © © © © © © © © © © © © © © © © © © © 3 *̂Comparison of
Foraging Patterns of Kestrels andShrxkes o o © © © © © © © © © © ©
© © © © © © © © © 43
Concurrent Goals © © © © © © © © © © © © © © © © © © o
44Conclusxons © o o © © © © © © © © © © © © © © © © © © 46
3© PATTERNS OF HUNTING WHILE HOVERING © © © © © © © © © © © ©
48Methods © © © © © © © © © © © © © © © © © © © © © ©• ©
49Advantages of Hunting While Hovering © © © © © © © © © 49Costs of
Hoverxng © © © © © © © © © © © © © © © © © © 31Hoverxng Hexght o o
© © © © © © © © © © © © © © © © © 39Optimal Wind Speed for Hovering
© © © © © © © © © © © 61Hoverxng Txme © © © o © © © © © © © © © © ©
© © © © © 66Hovering as an Alternate Hunting Technique © © © © © ©
68
LIST OF REFERENCES © © © © © © o © © © © © © © © © © © © ©
71
iv
-
LIST OF ILLUSTRATIONS
Figure Pagelo Geometric considerations of perches and vegetation
„ = = 82= Relative areas visible to a perched bird showing
effects
of grass height and density 0 = 0 0 0 0 0 0 = o b o o = 103=
Approximate increase in visible area with increasing
hunting height 0000. 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 114o
Approximate prey encounter rate with increasing perch
height o 0 0 0 0 0 o 0 0 - 0 0 0 00 00 o o o o 00 0 0 0 155o
Effects of increasing height on net energy gain per
attack 0 0 0 0 0 0 0 0 0 0 00 0 0 0 0 0 0 0 0 0 0 0 0 0 l660
Rate of net energy gain as a function of hunting height. 177o Index
of grasshopper abundance in months of August
through December 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 , 00 2^80
Success rates and lengths of giving-up times for shrikes
(A) and kestrels (B) as functions of season 0 0 0 0 0 0 339= Two
possible mechanisms for threshold renewal 0 0 0 0 0 4110o The
effect of air speed on the power required to fly 0 o 9211= Hovering
effort as a function of wind speed 0 = 0 0 = 0 9812= Hovering
height as a function of wind speed at 2. m = « = 69
v
-
LIST OF TABLES
Tablelo Effects of perch height on the distances traveled to
prey for kestrels and shrikes » o o . < , » . o o . . o o o2=
Perch height related to time of year o o =, 0 « 0 , o = o-3= The
relation between perch height and wind speed4.e. Effects of perch
height on distances traveled between
p e r c h e s o o o o o o o o o o o o o o o o o o o o o o o b
o
5® Effects of wind speed on distance traveled betweenperches 0 0
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
60 The effects of perch height on giving-up time forkestrels and
shi* ikze s o o o o o o o o 00 0.0.00 00 o o
7o Distances to prey at different times of year 0 0 o o 0080
Effects of distance to prey on success rates of kestrels
and shr r k e S o o o o o o o o o o 0 0 0 0 0 0 0 0 0 0 0 0 09 o
Effects of time on perch on success rates of kestrels
and shrikes 00 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 010o
Effects of time since last prey capture on success rate
of kestrels 00 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 011o
Comparison of prey types' and rates of prey capture from
perches and hovers 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 01 2 o
Calculations of Vmp for the American Kestrel o o o 0 0 o
VI
Page
132626
29
29
3236
38
38
39
5563
-
ABSTRACT
Although considerable literature on optimal foraging theory
exists, few field tests have been conducted* To make such tests,
winter foraging patterns of American Kestrels (Falco sparverius)
and loggerhead Shrikes (Lanins ludovicianus) were observed in
southeastern Arizona to compare actual patterns with predictions of
an optimal foraging model developed for predatory ground-hunting
birds* The model is developed from considerations of foraging
theory, energetics, and perch and vegetational characteristics that
influence vision of the predator* Two hunting techniques are
analyzed; hunting from perches by kestrels and shrikes, and hunting
while hovering by kestrels*
Analysis of hunting from perches includes patch selection,
movement between patches, allocation of time in patches, and prey
selection* For kestrels and shrikes, patch selection primarily
involves selection of a perch* Considerations of factors affecting
hunting from perches predict the existence of an optimal hunting
height which increases with decreasing prey abundance and
increasing prey size* When comparable prey decrease in abundance,
kestrels and shrikes hunt more often from higher perches* Selection
of perch height is also affected by wind; birds perch lower at high
wind velocities* Kestrels and shrikes appear to minimize time and
energy spent traveling between patches; they nearly always forage
unidirectionally and travel greater distances between high perches
than low ones* Givirig-up times, i*e*,
vii
-
times spent in patches where no prey were attacked, appear to be
determined in part by previous hunting times; giving-up time
correlated better with the previous three hunting times than just
the last one0 Prey selection appears to be strongly influenced by
three factors: distance from perch, evaluation of probability of
success, and size and type of preyo Success rate decreases with
hunting time„ The interpretation is that a threshold of prey
selectivity diminishes with timeo Such a diminishing threshold
could account for partial preferences in diets0 Contrary to
predictions of some optimal foraging models, prey selectivity
appeared to increase with decreasing prey densityo An explanation
for this pattern may be that birds minimize variance in food intake
by avoiding riskso
' Analysis of hunting while hovering primarily concerns the
energetics of hovering flight and their effects on the utilization
of this foraging method,. Hovering allows kestrels to hunt in areas
without suitable perches, but the relatively high energetic costs
restrict its use to times of favorable wind speeds,. The optimal
wind speed at which to hover is apparently equal to the air speed
at which flight is least costly= Most hovering occurs when optimal
wind speed and optimal hunting height coincide; when they do not,
kestrels appear to adopt a compromise between the two„ Because wind
speed increases with height, hovering height decreases as wind
speed increases« Duration of individual hovers from which prey was
not attacked was affected by time of year, duration of the previous
hover from which prey was attacked, and wind speed* Rate of energy
intake is greater when
-
hovering than when hunting from perches» Hovering appears to be
an important alternative foraging" strategy for some species of
birds at times of favorable environmental conditions^
-
CHAPTER 1
INTRODUCTION
Optimal foraging theory shows great promise for providing a
better understanding of animal behavior and community structure
(Pyke, Pulliam and Charnov 1977), but relatively few studies have
fully assessed its application in natural systems* In most papers,
Optimal foraging has been treated only theoretically on a strategic
level, e * go, Schoener 1971 and Charnov 1973= The scarcity of
field tests may be in part due to difficulties in translating
theory on a strategic level to testable predictions on a tactical
level* On a strategic level, terms are often vaguely defined and it
is possible to focus on only one variable while others are ignored*
On a tactical level, terms must be defined more precisely and many
variables that potentially affect an animal's behavior must be
considered simultaneously* Another problem that may contribute to
the scarcity of field tests of foraging theory is the difficulty in
selecting a system where an animal can be observed for extended
periods*
A crucial part of optimal foraging models is identification of
an animal's goal (Schoener 1971, Charnov 1973, Pyke et al* 1977)=
Although the choice of goal may affect the overall time budget of
an animal, many goals ultimately reduce to the prediction that an
animal
' should attempt to maximize net energy intake while foraging*
To do
1
-
this, an animal must make a number of choices, Charnov (1973)
identified a hierarchy of such choices: a habitat in which to hunt,
apatch within that habitat, a foraging method to use in the patch,
and prey types to be pursued. Although I believe that such a
hierarchy, is a useful tool for analyzing foraging behavior, I do
not believe that the four choices must occur in the order listed.
In particular, foraging method may be determined before habitat or
patch selection occurs because particular kinds of animals may be
constrained by evolutionary adaptations which restrict their range
of foraging methods.
In this paper I construct a tactical model for some aspects of
foraging of ground-hunting predatory birds from considerations of
perch and vegetation characteristics, energetic costs, and ideas
from optimal foraging literature. Foraging of these birds provides
a good system to test foraging theory because complicating
variables are minimized, terms can be operationally defined, and
foraging activities are easily observed. The model is developed
assuming that these birds are attempting to maximize net energy
intake while foraging. This goal appears to be appropriate for
predatory birds, and all foraging behaviors in this study were
predicted from this assumption. However, some data collected during
this study suggest that prey selection may also be influenced by
minimizing variance in energy intake. In most cases predictions of
foraging behavior generated from both goals are the same, and,
therefore, discrimination between the two is not usually critical.
Concurrent goals, such as avoiding predation, or maintaining
territories do not appear to significantly affect foraging behavior
of
-
these birds,, A more thorough discussion of these factors is
presented later in this paper»
Qualitative predictions of the model developed are tested and,
in many cases, verified in the field with foraging patterns of
American Kestrels (Falco sparverius) and Loggerhead Shrikes (Lanins
ludovicianusX Predictions are based primarily on foraging theory
and bonsiderations of flight energetics and geometric properties of
hunting from perches, but were also biased by known information of
kestrel biology0 Some predictions were changed during the course of
the study in light of new considerations, but all predictions were
a priori in the sense that they were made before the extensive data
were analysed«, These predictions can also be treated as hypotheses
and tested independently by other investigators working with other
organises or in different habitatso
Pyke et al= (1977) have divided foraging theory into four
categories: diet, patch choice, allocation of time in patches, and
patterns of movement between patcheso In Chapter 2 of this paper,
foraging method, hunting from perches, is treated as a constant
while behaviors associated with patch choice, allocation of time in
patches, and patterns of movement between patches are examinedo I
also analyze some aspects of diet, specifically quality evaluation
of prey by distance and capture success rate* In Chapter 3, I
analyze factors influencing the choice between two foraging methods
for ground-hunting predatory birds, hunting while hovering and
hunting from perches^Patch choice and allocation of time in patches
for birds hunting while hovering are also examined®
-
CHAPTER 2
PATTERNS OF HUNTING FROM PERCHES
An Equation for Net Energy GainRate of net energy gain of a bird
hunting from perches can be
represented by the equationsEg = E/A ° A/t - RMR - C/t (1)
where Eg is net rate of energy gain; E/A is the net energy
gained per attack; A/t is the attack rate; RMR is resting metabolic
rate, here defined as all the energy required to hunt from a perch
including thermoregulation; and C/t is the rate of energy expended
changing perches when no prey are attacked,,
Net energy gained per attack (E/A) is a function of other
variables such that:
where fs is the frequency of success (success rate), e is mean
energy content of prey attacked, and a is the mean energy expended
in making an attack including costs to fly to the ground and return
to a perch0
Similarly, attack rate (A/t) is a function of other variables'
such that:
where Pp is the proportion of prey encountered that are attacked
and N/t is the encounter rate with prey over the entire foraging
bouto
E/A = fs(e) - a (2)
A/t = Pp (N/t) (3)
-
• 5To increase net energy intake, a bird can increase E/A or
A/t
or decrease KMR or C/t, E/A can be increased by increasing fs,
or e, orby decreasing ao Attack rate can be increased by increasing
N/t or Ppo Because these variables are interrelated and tradeoffs
occur between some, the exact combination of values that results in
a maximum net energy gain depends on the relative values of eacho
Because many of the terms cannot be measured, these equations will
not be evaluated numerically but used to provide an understanding
of the factors that affect hunting from percheso Qualitative
predictions and analyses of foraging behaviors can then be
madeo
MethodsObservations of foraging kestrels and shrikes were made
in the
grasslands of southeastern Arizona from September 1975 to March
1977° Although some data were collected throughout the year, most
observations were made in fall and winter monthso Most data were
collectedbetween 0900 and 1500 h0 Birds were watched with lOx
binoculars or a 15-60x telescope from a parked vehicle0
Data taken on foraging birds included perch height, distance
traveled between perches, distance to prey, success rate of
attacks, and time spent hunting on percheso A bird was considered
to be foraging when it showed active signs of searching the groundo
Except for a few times in early fall, birds appeared to forage
almost constantly„ Time spent in nonforaging activities (such as
preening) was subtracted from the time on perches* Most birds were
followed as long as possible *
-
Times were measured with a stopwatch and data were recorded on a
portable tape recorder and transcribed later. Heights and distances
were estimated visually but were calibrated periodically by taking
precise measurements. Wind speeds were measured with a Dwyer
handheld wind meter.
In one area perches consisting of poles (agave stalks) 3 to 5 m
high were erected on three successive fenceposts spaced 3 m apart
such that perch height increased from 2 (fenceposts) to 5 m at
approximately one meter intervals. Only 2 such units were erected,
but 13 others of perches 2, 3, and 4 m high and 4 units of 2 and 3
m poles were also constructed serially in the same area.
No quantitative study of prey populations was conducted but
grasshoppers were censused along a 1750 m route in grassland
habitat. Kestrel diets were monitored by analysis of pellets found
at roosts.
Patch ChoiceAlthough the term "patch" has been widely used in
the litera
ture, it is often ambiguous and poorly defined. For perch
hunting birds, a patch can be operationally defined as the area
that can be hunted from a perch; thus, time in a patch and movement
between patches are easily measured.
Net caloric intake can be increased by foraging in patches where
encounter rate with prey (N/t) is high. For birds hunting from
perches, encounter rate is a function of prey availability and area
hunted. Prey availability is some function of prey density, prey
type, vegetational structure, and weather. One way encounter rate
can be
-
increased is by hunting in areas where prey availability is
highero In a fine-grained situation, patches must be visited for
prey availability to be assessed^ Thus, variations in prey
availability would have little effect on patch selection, although
it would contribute significantly to habitat selection* Much of the
area in which kestrels and shrikes foraged appeared to be
homogeneous so that birds probably could not assess prey
availability before visiting patches*
Encounter rate can also be increased by hunting a larger area*
Area hunted can be increased by hunting in habitats with little
vegetation so visibility is increased, and by using higher perches.
But increasing perch height also increases foraging costs and
handling time of prey. The following analysis of the geometry of
perches and vegetative structure on the terms of equations (1) and
(2) suggests that there exists an optimal height from which to hunt
arid that patches should be chosen on the basis of perch height and
vegetative structure.
Geometry of Hunting from PerchesFigure 1 is a model that
provides a basis for estimating the
relative area of ground that is visible from a perch, where h
equals perch height, g is the. average height of grass clumps or
other vegetation, d is the average distance between these clumps,
and y is the distance from a given clump to the base of the perch
(y is a multiple of d). There is a distance, x, behind each clump
where the ground is not visible from the top of the perch. This
distance increases with increasing distance of the clump from the
perch until at some point it equals the average distance between
grass clumps and no ground is
-
8
\\\
h
\ \\\
Figure 1. Geometric considerations of perches and
vegetation,
-
9visible« Thus, ground area visible to a perched bird can be
visualized as concentric rings of decreasing width around a percho
Though the width of each ring decreases with distance, the size
increases so that the area of each ring does not necessarily
decrease„ Ring area as a function of distance from the perch,
depends on the average clump distance and height but, in general,
increases to a point and then decreases = Some examples are shown
in Figure 2o When grass clumps are tall and closely spaced, very
little ground is visible0
By increasing perch height, a bird can hunt more area because
the distance behind each grass clump that is not visible
decreases»But the geometric properties are such that increasing
increments of perch heights result in successively small decreases
in x For . the area within a given radius around a perch, area
that. can be hunted increases in the manner shown in Figure J>
and becomes asymptotic at the maximum area within the specified
radiuso
This development of effects of perch height is based on
simplified but robust assumptions«, Grass clumps or other
vegetation obviously are not opaque and of even height, and do not
occur in continuous concentric rings at regular distances around
perches= The following analysis also assumes that prey are flato
However, considerations of the real properties of vegetation and
prey have little effect oh the qualitative aspects of the model
which realistically indicates the unavailability of some prey in
vegetation*
This analysis of perch geometry and area of the ground visible
from perches leads to the following prediction*
-
VISI
BLE
AR
EA
DISTANCE FROM PERCH
Figure 2. Relative areas visible to a perched bird showing
effects of grass height and density. — For curves A and B, g = 50
(tall grass) and d = 8 and 12* respectively. For curves C and D, g
= 5 (short grass) and d = 8 and 12* respectively. Areas were
calculated on the basis of a perch height (h) equal to 900. All
numbers in cm.
HO
-
VISI
BLE
AR
EA
11
HEIGHT
Figure 3* Approximate increase in visible area with
increasinghunting height. — Dotted line represents maximum area
visible within a specified radius around a perch (see text).
-
12Prediction 1: A greater proportion of attacks should occur
at
greater distances from tall perches than from short ones because
more area is visible at greater distances,.
Test of Prediction Is Kestrels and shrikes made greater
proportions of attacks at greater distances from higher perches
(Table 1)„ Distances traveled for prey were significantly shorter
for shrikesthan for kestrels from perches of equal heights (for
perches
If prey are taken primarily from the ground, which appears to be
a valid assumption for kestrels and shrikes, rate of prey
encountered per search time (N/tg) should increase with height in
approximately the same manner as area that can be hunted (Figo 3)°
But handling time
-
13Table l o Effects of perch height on the distances traveled to
prey
for kestrels and shrikes»
Kestrels ShrikesPerch % attacks at: attacks at:
Height (m) n 0-20 m 21-40 m >40 m n 0-10 m 11-20 m >20
m
-
also increases with height because the time to attack and return
to a perch increases* As handling time increases, search time
decreases; thus, encounter rate for the total time hunting (N/t)
increases with perch height to a maximum and then decreases, as
shown in Figure 4*The exact shape of the curve depends on the
relative values of N/tg and th/tSo Increasing prey density
increases encounter rate per search time (N/tg) but does not affect
handling time; therefore, the perch height where encounter rate is
maximized decreases as prey density increases*
Increases in perch height also increase foraging costs* Cost to
attack prey (a) increases with height because the cost to return to
the perch increases, though the cost of the drop from the perch to
the ground is probably negligible because it is gravity assisted*
It seems reasonable that cost of an attack is directly proportional
to height* The increased height causes the energy gained per attack
(E/A) in Equation (1) to decrease as shown in Figure 5° If mean
energy content of prey were increased, the line in Figure 5 would
shift'Upwards*
If attack rate were proportional to encounter rate arid net
energy gain per attack decreased with height as outlined above, an
optimal hunting height, where the net rate on energy gained is
maximized, could be found by multiplying the equations of the
curves in Figures 4 and 5* The result of such a multiplication is
shown in Figure 6*. The preceding analysis suggests that optimal
hunting height increases as mean prey size increases or as density
decreases*
In some cases, resting metabolic rate might have a significant
effect on optimal hunting height* RMR varies with environmental
-
PREY
EN
COUN
TER
RA
TE15
HEIGHT
Figure 4. Approximate prey encounter rate with increasing perch
height.
-
NET
ENER
GY
GAIN
PE
R A
TTA
CK
16
HEIGHT
Figure 5* Effects of increasing height on net energy gain per
attack. — Line B is for larger prey.
-
RATE
OF
NE
T EN
ERGY
G
AIN
17
HEIGHT
Figure 6. Rate of net energy gain as a function of hunting
height. — This curve is obtained by multiplying net energy gain per
attack (Fig. 5) and attack rate, which is assumed to be
proportional to encounter rate (Fig. 4). See text for further
explanation.
-
18conditions, especially temperature and windo At times of high
winds, RMR could increase due to heat loss or an increase in the
effort required to remain on a percho Because wind speed increases
with height, BMB should be greater on higher percheso Birds could
reduce this cost by perching lower or in a more protected place,
otherwise RMR is a fixed cost for any given time or place
-
19increase in cost is probably small in comparison to costs of
making attacks because only a horizontal flight is required*
Predictions and TestsThe previous discussion of the factors
influencing costs and
benefits of foraging from perches suggests that patch selection
should be based on perch height and vegetation density* From this
analysis I make the following predictions*
Prediction 2s Areas of tall, dense vegetation should be avoided
because little ground is visible regardless of perch height and the
probability of prey escaping in the vegetation is high*
Prediction 3s Because larger birds generally take larger prey
than smaller ones, their optimal hunting height should be higher
and they should select higher perches* Different-sized birds are
not strictly comparable, however, because energy to gain height is
not the sane * Nevertheless , female kestrels, which weigh about
110 g, would be expected to perch the highest, male kestrels (about
100 g) slightly lower, and shrikes (about 50 g) considerably lower
than either sex of kestrel*
Prediction 4: Optimal hunting height should increase as mean
preysize increases or as prey density decreases*
Prediction 5: If wind velocity is sufficient to increase
energetic costs due to heat loss or effort to remain on perches,
optimal hunting height should decrease with increasing wind speed
because wind speed is lower near the ground*
-
20Testing these predictions in the field was complicated by
several factors,. Perches in nature rarely present birds with
continuous choices of height« In the study area, fenceposts (lo5 to
2 m) and utility poles and wires (8 to 10 m) were the most common
and often the only perches available„ Some perches were apparently
not suitable for reasons other than height,. Neither kestrels nor
shrikes were ever seen perched on electric wires of utility poles;
telephone wires were always usedo Both species also showed a
definite preference for perches that provided greater stability;
wooden fenceposts were preferred to metal ones, utility poles or
wires near poles were preferred to wires midway between poles,.
Because these respective perches were usually close in proximity
and of similar height, however, these preferences had little
influence on perch height selection,,
Test of Prediction 2: Kestrels and shrikes clearly avoided
hunting in areas of tall, dense vegetation,. During the months when
grasshoppers were abundant and were the primary food, kestrels and
shrikes were observed hunting only in areas of short grass even
though grasshoppers appeared to be more abundant in areas with tall
grasso Avoidance of areas of tall, dense vegetation was best
demonstrated by several observations of kestrels foraging
sequentially along utility wires that crossed an area of tall,
dense grass (Sporobolus wrightii) bordered by areas of short,
sparse grasso Upon reaching the area of tall grass after foraging
in the area of short grass, kestrels made flights much longer than
the usual distance between hunting perches across the tall grass
and resumed foraging in the area of short grass on the other
side,.
-
21Hunting in areas of tall, dense grass might be profitable
if
higher prey availability or greater prey size compensated for
the low visibility0 Observations of a male kestrel hunting in a
small clearing in tall, dense grass in February, when insect prey
were scarce, suggested that such compensation may sometimes occur»
After making a number of aborted attacks near the edges of the tall
grass, a cotton rat (Sigmodon sp0) was capturedo Cotton rats are
among the largest prey items that I recorded in the diets of
kestrels in southeastern Arizona and were very abundant in the tall
grass areas that winter«,
This observation also provided a possible example of hunting
height being affected by a decreased probability of success with an
increased distance from prey0 It seems reasonable that cotton rats
were exposed to capture only at the edges of the tall grass for
short periods of time» In order for an attack to be successful, the
kestrel would have to perch a short distance away to reduce the
time to reach the preyo Even though utility wires were available
nearby, the kestrel hunted only from perches barely higher than the
surrounding grass (t 105 m)o Such a reduction in hunting height is
profitable only if encounter rate is high or prey size is
large=
Test of Prediction J>% Mean perch height was highest for
female kestrels (7°5 m), intermediate for male kestrels (6=3 m),
and lowest for shrikes (5=0 m)» Because perches were normally
either fenceposts or utility lines, perch differences are perhaps
best shown by the percentage of times the birds perched higher than
?06 m (25 ft0)0 Kestrels perched at heights of 7=6 m or higher
significantly more often (53̂ 9 n = 602) than did shrikes (3^, n =
217; = 23=9, p < o001),
-
22but there was no significant difference between male ( , n =
11?) and
pfemale kestrels (52^, n = 485; ̂ = 069, p > o90)= These
patterns areas expected for reasons of optimal hunting height, but
I have other data which suggest that perch selection also was
affected by aggressive interactionso
The most striking characteristic of perched kestrels was that
they chose the top of the tallest available perches* Ninety-three
percent (n = 688) of kestrels observed were on the tallest perches
available within a distance of 25 m* Shrikes also perched
frequently on the highest perches but they did so a smaller
percentage of the time (85#, n = 359) Selection of the tallest
perches was most clearly demonstrated from observations of birds on
manipulated perches. Both kestrels and shrikes always chose the
highest pole in a set (n = 14 and 40, respectively) even though the
highest poles were of different absolute heights in different
sets.
In many areas differences between the highest and lowest perches
were substantial, and perches of intermediate height were not
available, but even in areas where intermediate perches were
available, the highest perches were chosen. Where telephone wires
ranged from heights of 6 to 9 m, the highest were chosen. Likewise,
in leafless trees where an almost continuous range of perch heights
was available, kestrels and shrikes nearly always perched within 1
m from the top on the highest good-sized branch. Large leafy trees
presented an interesting situation. To maximize area hunted, a bird
should perch on the side of the tree because from the top the view
of the ground below would be blocked by the foliage. Kestrels
consistently perched on the
-
23sides of these trees rather than at the top0 Shrikes were not
observed in these treeso
These observations suggest that optimal hunting height for
kestrels was usually higher than available perches® This is also
suggested by the heights of birds using an alternative hunting
techniques hoveringo Hovering height was usually between 1101 and
l4®3 m (see Chapter 3), higher than virtually all perches on the
study area®
Test of Prediction 4: It is difficult to assess whether changes
inprey size or density affected perch height becauses (1) optimal
hunting height often appeared to be higher than available perches;
(2) it was difficult to assess changes in prey sizes and densities;
and (3) wind was a confounding variable = It is also possible that
optimal hunting height is primarily determined by the largest prey
if these account for a large proportion of the prey biomass®
However, field observations indicated that from August through
December diets of kestrels and shrikes consisted primarily of
grasshoppers® Grasshopper populations showed a marked decrease
during this time (Fig® 7)® Although the grasshopper population
Consisted of individuals of many body sizes much of the year, most
seen after August were large ( >2®5 cm) and from September
through December no marked change in their size was apparent®
Analysis of kestrel pellets showed that the diet contained more
rodents toward the end of this period® This decrease in prey
density for both kestrels and shrikes and the inclusion of more
rodents in kestrel diets should cause an increase in optimal
hunting height® Kestrels and shrikes perched more often on perches
>8 m on days of wind speeds
-
NUM
BER
OF
GR
ASS
HO
PPER
S
2500
2000
1500
1000
500
15 AUG I SEP 15 SEP I OCT 15 OCT I NOV 15 NOV I DEC 15
DECDATE
Figure ?• Index of grasshopper abundance in months of August
through December. — Points indicate census dates.
-
25(Table 2) => Differences are significant; for kestrels, =
8o3? P < =005; for shrikes, X = 19=0, p < o005o
Other evidence for changes in hunting height due to changes in
prey density and size comes from data on hovering kestrels (Chapter
3) o On several occasions kestrels hovering lower than usual were
observed apparently capturing small abundant prey iterns0 Also,
when no attacks were made on prey, successive hovers tended to be
at increased heights Suggesting that the birds' estimates of prey
densities decreased and hunting height was adjusted accordingly„
This may also explain observations by Pinkowski (1977) that
bluebirds (Sialia sialis) moved to a
2higher perch (n = 65) significantly more often (X = 8o3? p
-
26Table 2= Perch height related to time of year=
Time Period Times Perched8 m
Kestrels 1 Septo=17 Octo 39 4530 Octo~31 Deco 20 60
Shrikes 1 Septo-17 Octo 23 530 Octo-31 Deco 23 46
Table 3o The relation between perch height and wind speedo
Wind Speed (mph) Times Seen at Perch Height0=3 m 4-7 m >8
m
Kestrels 10 116 33 96
Shrikes 10 100 13 7
-
Movement Between Patches In addition to choosing patches,
foraging animals must make
decisions about moving between patcheso In many cases movement
between patches is very complex because of the multidimensional
nature and the effects of patch boundaries (e0go, Pyke 1978)0
Probably for these reasons few predictions or tests concerning
movement between patches have appeared, though Gharnov (1973) has
discussed some theoretical aspects of this topic and suggested that
prey distribution is an important factoro For kestrels and shrikes,
hunting from utility lines or fences in fairly homogeneous
grasslands, movements between patches are limited to one dimension;
thus, aspects of between-patch movement are simplified» Choices
concerning movements between patches are restricted to whether to
return to the same perch after an attack, which direction to go to
the next perch, and how far to move* Here I consider only the
latter two choices; the decision whether to return or not is
apparently complicated and will be discussed elsewhere» However,
kestrels and shrikes usually did riot return to the same perch
after an attempt for preye
Net energy intake can be increased by decreasing the time and
energy spent traveling between patches (C of Equation 1)® For birds
hunting from a line of continuous perches, I make the following
prediction®
Prediction 6s Kestrels and shrikes should forage
unidirectionally and should move only far enough between patches so
that overlap with adjacent patches is minimal® Due to difficulties
in calculating the area that can be seen from a perch and the
difficulties measuring
-
28appropriate parameters in the field, I cannot predict actual
distances between patches,. However, a qualitative prediction that
can be made is that distance between perches should be greater from
higher perches than low ones because more area is visible from each
percho
Test of Prediction 6: Kestrels and shrikes nearly always
foragedunidirectionaily along a line of continuous percheso Only
occasionally did a bird return to a perch after a visit to a
different one„ Distances between perches were significantly greater
from tall perches than short ones when kestrels and shrikes left
without attacking prey (Table 4; t = 3o5» P < =005; t = 6,1,
p
-
29Table 40 Effects of perch height on distances traveled
between
percheso — Only distances between continuous equal-height
perches after giving-up times are includedo
Perch Height (m) x Distance (m)» Kestrels x Distance Cm),
Shrikes
2-3 18*2 15=5>7 53=9 68 oO
Table % Effects of wind speed on distance traveled between
perches*
x Distance (m) Between Perches for:Perch Kestrels Shrikes
Height (m) Wind 10 mph Wind 10 mph
2-3 23°2 13o4 13o9 13=2>7 67=5 35=1 38=9
-
30tendency was noted for shrikes on low perches to hunt into the
wind at times of high wind speeds, and there was no difference in
distance between these perches for times of high and low wind (t =
37* p >o30)o Sample size of shrikes for distance between perches
on high perches was too small for analysis0
Allocation of Time in Patches Most studies of optimal allocation
of time in patches concern
"giving-up timeso” Giving-up time is the period waited since the
last capture before an animal leaves a patch* Although there is
general agreement that giving-up times are derived from information
from previous experience, the kind and quality of information
animals use has not been determined* Charnov (1973) has proposed
the marginal-value theorem, a deterministic model that relies only
on the mean times waited in previous patches* This model has
recently been criticized by Oaten (1977), who suggested that a
stochastic model, where an animal uses the variance as well as the
means, is necessary for optimal foraging* It also seems likely that
information gathered while forage ing in a patch may affect
giving-up time* For birds hunting from perches, one such source of
information may be assessment of prey that are seen but not
attacked*
-Qiving-up time could be measured for kestrels and shrikes when
they left a patch without attacking prey* My limited data on
kestrels and shrikes does not allow a determination of exactly how
these birds use past experience to determine giving-up time*
However, it seems that part of the information used should be the
means of some number
-
of times waited in previous patches before prey were attacked,.
Therefore, ’I make the following prediction*
Prediction 7: Giving-up time should correlate with some number
ofprevious times waited for prey*
Test of Prediction 7° The mean of the last three times waited in
a patch before prey was attacked was a better predictor of
giving-up time than just the last time* For kestrels r = =59 (n
=17, p < o005) and r = .14 (n = 32, p >.25), respectively;
for shrikes r = .78 (n =32, p
-
32Table 6= The effects of perch height on giving-up times for
kestrels
and shrikeSo
Perch Height (m) h* x Giving-up Time (s)
Kestrels 0-3 34 146 o 634 18606
Shrikes 0-3 72 67=7>3 13 l83o4
Includes only giving-up times of less than 600 s0
-
33
Figure 8
75
65
5 55
i45if)if)LU8 35 Z)if)
75
65
55
45
35
LxJ
crif)if)LUOUz>if)
SO ND JF MB
95 <I
85 iif)
75 LU
65 z>io55 5
50 lx
280 «I I4if)
LU
SO ND JF M TIME PERIOD (MONTHS)
250
220 f-
190 %io160 >
oo |x
Success rates and lengths of giving-up times for shrikes (A) and
kestrels (B) as functions of season.
-
34is not unexpected because diets of both species varied during
the year. If diets shifted to smaller more abundant prey items as a
preferred prey decreased, giving-up. time would likely decrease
even though overall prey quality decreased. This appeared to be the
case for shrikes in. January and February; smaller prey were taken
and giving-up time decreased.
Optimal DietOptimal diet theory assumes that animals evaluate
prey and make
decisions whether or not to attack each item encountered. For
birds hunting from perches, evaluation can be based on prey size
(e), energy and time required to attack (a), or the birds'
estimates of chances of success (fs). Time and energy to attack
increase with and are affected primarily by distance to prey. As
previously discussed, success rate might also decrease as distance
to prey increases. Evaluation for chance of success seems
especially likely because of the high cost of an unsuccessful
attack. Any evaluation of prey and subsequent selectivity lowers
the proportion of prey attacked (Pp). From my observations, I am
able to examine aspects of prey selection based on distance to prey
and chance of success.
I am certain that kestrels evaluate prey. Hunting birds often
showed evidence of sighting prey with behaviors normally associated
with attacks such as head-bobs, tail jerks, and plumage depression,
and then did not attack. Shrikes showed similar behavior but less
obviously.
If kestrels and shrikes do not evaluate prey on the basis of
distance, the proportion of attacks made at any distance from a
perch
-
35should be proportional to the visible area at that distance= I
cannot evaluate whether this occurs because it requires a
quantitative measure of ground area that is visible as a function
of distance from the perch* Topographic irregularities and
vegetation opacity affect areas that can be hunted and are
difficult to measure in the field* Even if these problems are
neglected, quantitative evaluation of even the simple model
presented earlier is too complicated to be practical
-
36Table 7= Distances to prey at different times of yearo
From P e r c h e s m High From PerchesJ>8 m Hi^h Time x
Distance x Distance
Period n to Prey (m) n to Prey Cm)
Kestrels
1 Septo- 17 0cto30 Octo—31 Deco
32
10
1106
I606
31
kl
25=2
50=0
Shrikes1 Septo™ 17 Octo30 Octo«-31 Deco
22
19
5=2
11 o2
3
32
15=023=0
-
37Success rate did not decrease significantly with
increasing
pdistance- of attack for kestrels or shrikes (Table 8? X = 6=2,
p > o10;2
X = o20, p o95» respectively)c The apparently lower success rate
for. kestrels at very great distances is due almost entirely to
attacks on birds (11 of l6)» Either the probability of success did
not decrease with distance for the majority of prey or kestrels and
shrikes evaluated their chance of success and attacked only more
vulnerable prey at greater distances= Laboratory work by Sparrowe
(1972) showed that attack responses by kestrels were affected by
prey exposure time0 This suggests that evaluation of probable
success is at least partly responsible for the constant success
rate with distance0
Evidence that prey are evaluated on the chances of success is
that success rates of kestrels decreased significantly with hunting
time on a perch (Table 9| X^ = 15=59 p p >=10)o The last two
time categories for shrikes were combined for analysis0 Distance to
prey did not increase significantly with hunting time for kestrels
or shrikes (t = =62, p > 025? t = =29, p> o40, respectively)
o This pattern suggests that prey items are evaluated on the basis
of chance of success, and the threshold for an attack diminishes
with time= This threshold apparently is renewed when birds change
patches= The exact manner by which the threshold is renewed cannot
be determined from my data because success rate also decreased
significantly (X^ = 12=59 p < =005) with time since the last
capture for kestrels (Table 10; small samples precluded analysis
for
-
38Table 80 Effects of distance to prey on success rates of
kestrels and
shrikeso
0-20 Distance to Prey (m) 21—40 41-60 >6oattempts successful
66 24 10 8
Kestrels attempts unsuccessful 44 16 6 16% success 60 60 65
33
0-10 11-20 >20 'attempts successful 28 12 6
Shrikes attempts unsuccessful 21 8 3% success 57 60 67
Table 9° Effects of time on perch shrikes300
attempts successful 50 21 15 . 7Kestrels attempts unsuccessful
25 16 13 22
% success 67 57 54 24
0-4o 41-120 > 120attempts successful 19 18 8
Shrikes attempts unsuccessful 12 13 13% success 61 58 38
-
39Table 10= Effects of time since last prey capture on success
rate of
kestrelso
Time Since Last Capture (s) 0-300 301-600 >600
attempts successful attempts unsuccessful % success
408
83
8753
181949
-
4oshrikeso Renewal may be complete (Fig* 9a) or only partial
(Figo 9b) with each change of perch0
The diminishing threshold model of evaluation of capture success
provides a simple mechanism for partial preferences in diets if
assessment of prey types changes on a short time scale in a manner
similar to the chance of successo Most theories of optimal diet
predict that animals should not show partial preferences? i0e0? a
prey type should either be included in the diet every time it is
encountered or not at alio Pulliam (1974) has suggested that
partial preferences would be expected if dietary constraints were
important or if the predator’s assessment of prey densities changed
during the time it searched for prey« In a later review (Pyke et
alo 1977), dietary constraints are discussed at some length but
short-term assessment is not mentioned* The diminishing threshold
shown by kestrels and shrikes support the latter theory and suggest
that dietary constraints may not be necessary to explain partial
preferenceso
Kestrels also appear to evaluate escape strategies of prey0
Roest (1957) and Collapy (1973) have mentioned differences in
attack behavior for different prey types* These were also noted in
this study* For insects, kestrels usually glided down from a perch
with few wing- beats; for attacks on rodents and lizards, flights
from perches were usually direct with many wingbeats as if to
minimize time to reach the prey; for birds, attack flights were
fast and powered but kestrels dropped quickly from the perch and
completed the attack from grasstop level* The latter method
suggests that surprise is important when birds are attacked* .
'
-
THRE
SHO
LD
OF
SE
LEC
TIV
ITY
a perch change▲ perch change after prey capture
TIME
Figure 9« Two possible mechanisms for threshold renewal. — In
’’A*' threshold renewal occurs with each perch change regardless of
prey capture; in "B” the threshold is only partially renewed with
each perch change and completely renewed only after a prey
capture.
-
42The above discussion suggests that birds can control their
success rate by varying their threshold of selectivity«, One
factor that influences this threshold is prey size„ If prey are
small relative to the size of the predator, success rate must be
high to forage profitably, especially if the cost to attack each
prey is higho If prey are large, a lower success rstte may be
toleratedo
Some data suggest that success rate may be affected by an
aversion to the risk of starvation or falling below a positive
energybalance rather than simply maximizing net energy: gaino
Figure 8 showshow success rates of kestrels and shrikes covaried
with the lengths of time waited in patches where no prey were
attacked (giving-up times)o If lengths of giving-up times are
inversely proportional to prey densities, as suggested by Charnov
(1975)» these data indicate thatselectivity based on estimates of
chances of success increases as prey density decreaseso Craig
(1978) presents data for shrikes, showing asimilar relationship
between prey density and success rate® Thispattern conflicts with
optimal foraging theories predicting selectivity should decrease as
prey density decreaseso An explanation for this pattern may be that
when prey are scarce, birds minimize variance in food intake by
attacking only prey that have a high probability of capture, even
if such behavior may also lower the mean net energetic gaino In
this way risk of starvation decreaseso As food becomes less
plentiful and the probability of starvation increases, risk
aversion increaseso This seems especially likely for selectivity
based on chance of success because of the high cost of an
unsuccessful attacke If prey reach a critically low level, this
conservative strategy may
-
not be sufficient to provide the food requirements of the
animalP At such times, birds may be forced to take more risks and
attack prey with a low probability of capture success but a high
energetic reward
-
Concurrent GoalsAs outlined in the introduction, all predictions
were made
assuming the goal of maximizing net .energy, reward while
foragingo This goal seems reasonable for many animals (see Schoener
1971; Charnov 1973; Pyke et alo 1977), and the agreement between
predicted and observed foraging behaviors suggests it is
appropriate for kestrels and shrikes in winter0 Some aspects of
prey selection, however, appear to be influenced by risk avoidance
= In some systems, other goals such as escaping predators,
searching for mates, maximizing a specific component of the diet,
or territoriality, may operate concurrently and influence foraging
behavior0 I do not believe that any of these significantly
influenced the aspects of foraging behavior discussed in this
papero
Of the concurrent goals that might influence foraging behavior
of kestrels and shrikes, territoriality appears to be most likely0
Both species are territorial in winter (Cade 1955, Mills 1975,
Miller cited in Bent 1950)o One might argue that kestrels choose
the tallest perches to "advertize" territories or to better survey
territories for intruderso Unidirectional foraging may be a
mechanism to patrol territory boundaries,. But some patterns are
not consistent with goals of territorial defense* Shrikes do not
always perch on the highest perches; kestrels perch on the sides of
leafy trees where the area that can be hunted is maximized, not at
the top where intruders are more easily located* Behaviors
associated with boundary conflicts suggested that kestrel foraging
behavior was little influenced by territoriality* Birds at
territorial boundaries appeared to forage no
-
k5
differently than others, even when a neighbor was nearbyQ Very
little time was spent in territorial interactions and rarely did
birds fly long distances to pursue an intruder» In the boundary
disputes I observed, an intruding bird was attacked only when it
flew off a perch after prey= Neither bird involved appeared to
notice the other until movement occurred,. Cade (1955) and Welty
(1962) also have noted that movement of an intruder is often
necessary to elicit an attack from a kestrel= This seems a
reasonable method to defend a feeding territory at relatively low
costo An intruder is no detriment as long as it takes no prey from
the territoryo If an intruder is prevented from capturing prey, it
will be advantageous for it to forage elsewhere0
Although kestrels and shrikes are occasionally preyed upon by
other raptors, it is apparently rare0 During this study the only,
attack on a kestrel that I witnessed was an unsuccessful one by a
Cooper* s Hawk (Accipiter cooperi)o This attack occurred in an area
of fairly dense oak woodland? no Cooper*s Hawks were seen in the
open grasslandso Kestrels showed little concern for other raptors
except to occasionally mob a Red-tailed Hawk (Buteo jamaicensus) or
Prairie Falcon (Falco mexicanus)o One shrike showed some alarm when
a Marsh Hawk (Circus cyaneus) passed near but, except for attacks
by kestrels which appeared to be motivated by competition rather
than predation, no attacks on shrikes were observed,.
Most data were taken at a time of year when searching for mates
was evidently of little importance= Some kestrels remained paired
in winter; these birds appeared to forage no differently than
unpaired one So
-
46Although some particular component of the diet may be an
espe
cially important requirement for some species, it seems unlikely
that carnivorous animals would have to take certain prey types in
order to obtain essential nutrients. Even if this were the case,
the searching behaviors studied here would be little affected. At
times, however, kestrels appeared to search for a specific prey
type. In addition to a kestrel apparently hunting Sigmodon in tall
grass, on at least two other occasions it appeared that rodents
were being hunted specifically. In these cases, kestrels hunted
small areas for long periods. It appeared that a rodent had been
sighted there previously and the kestrel was waiting for it to
reappear.
ConclusionsForaging patterns of kestrels and shrikes are
consistent with
predictions of a tactical model for ground-hunting raptors
developed from considerations of perch geometry and optimal
foraging theory.These patterns show that kestrels and shrikes can
measure distance and time and respond appropriately to quantities
such as means and, perhaps, variances. These are not unexpected
results. Perhaps more important than demonstrating that animals
appear to be selected to optimize foraging behavior is the
demonstration of the uses of optimal foraging theory as a tool to
better understand animal behavior. Optimal foraging theory is
certainly useful in understanding and examining decisionmaking
processes that enable animals to solve problems posed by
alternative prey types with variable temporal and spatial
distributions. It also shows great promise in analyzing and
understanding community
-
structure0 I am currently working on a paper which uses optimal
foraging theory to analyze coexistence among predatory birdso
-
CHAPTER 3
PATTERNS OF HUNTING WHILE HOVERING
Many animals, especially predators, have a repertoire of
foraging techniques. An animal would be expected to use the
technique that best meets its foraging goal, which often may be to
maximize net energy gain while foraging (Pyke et al. 1977)° Various
techniques are often used to take different prey types but may also
be used by one animal to hunt one prey type. When different
techniques are used to hunt a single prey type, variations in prey
behavior or environmental conditions are apparently responsible for
the choice of technique. In this chapter I examine hovering, a
hunting technique used by some birds, and some factors that affect
its use.
Although hovering flight is possible for most birds and many
species occasionally hover momentarily, only a few species hover
habitually while foraging. Aside from hummingbirds, regular
hovering is fairly restricted to predatory species that hunt over
open ground or water and take prey from the surface. Physiological
aspects of hovering flight of hummingbirds (Weis-Fogh 1972, 1973)
and ecological aspects of hovering by Ospreys (Pandion haliatus)
(Grubb 1977) have been studied, but ecological aspects of hovering
of other species havebeen studied only casually (Roest 1957;
Balgooyen 1976). Here, I/analyze patterns of hovering in the
American Kestrel (Falco sparverius) and relate them to a model of
optimal foraging.
48
-
49Methods
I collected data on foraging of American Kestrels in the
grasslands of southeastern Arizona from 1975 to 1977= Most field
work was conducted in fall and winter months= Data collected
included hovering heights, hovering times, times on perches,
attempts to capture prey and their success, and types of prey taken
from perches and hoverso Hovering time was determined by recording
the start and finish of a hover on a portable tape recorder and
then timing it later» The recorder was found to be accurate to
within less than d sec/min0 Wind speeds were measured at a height
of 2 m with a Dwyer hand-held wind gauge and hovering heights were
estimated visually by comparison with stationary objectso A
Dietzgen "Duo-site Range Height Finder" was also used to measure
hovering height but was found to be of little value due to large
distances between the kestrels and the observero When kestrels were
close enough for the instrument to be useful, results agreed well
with my estimates*
Advantages of Hunting While Hovering Although a small percentage
of the hovers observed were clearly
over prey that had been previously located from perches, hovers
were primarily used to search for prey* This was clearly
demonstrated by the systematic manner in which hovering occurred*
Two common patterns of hovering were series of hovers en route from
one line of utility poles to another and series of hovers parallel
to a line of utility poles made after flying some distance
perpendicular to these poles and then returning to the same line
some distance away* In these series distances between hovers were
fairly constant*
-
50Occasionally, it appeared that some hovers were made after
prey
had been located from a previous hover= In these cases, there
was essentially no horizontal distance between successive hovers;
one or more subsequent hovers were made after rapid, nearly
vertical drops as described by Roest (1957)° Seventy-three percent
(11 of 15) of these series resulted in attacks whereas attempts for
prey were made from only 29^ (126 of *+30) of all hovers
recorded,.
Many birds that habitually hover also hunt from perches0
However, areas that can be hunted from perches are limited by the
distributions and heights of those percheSo Advantages for hovering
appear to be that the rate of prey capture can be increased by
hunting areas where prey are more available and that optimal
hunting height can be more closely approachedo In Chapter 2, I have
demonstrated the existence of such optimal hunting height0 I would
then predict the following,.
Prediction It Hovering should not occur close to perches of
heights similar to hovering heightso
Test of Prediction 1: Of 402 hovers observed where distances
toutility wires were determined, over 94% were made at distances
greater than 40 m from those wires,. Hovers that were made within
40 m were low and momentary (
-
51telephone lines approximately 9 m higho Hovering height varied
with wind speed, but most (71%) hovers occurred above 9 mo Some
hovers were made above fences less than 2 m high that were also
used as hunting perches by kestrels, but these hovers were nearly
always above 9 m0
Costs of Hovering Hovering is energetically more costly than
hunting from a
percho For hovering to be advantageous, it should be used only
when the increased cost is exceeded by an increased energy rewardo
For hummingbirds the magnitude of the increased cost of hovering
over perching is reasonably treated as a constant, but for most
birds the cost of hovering is affected by wind speedo Hovering is
flying at zero ground speed, so that as wind speed increases a
hovering bird must increase its air speed to maintain zero ground
speedo (Note that unless a bird can fly backwards or sideways it
must face the Wind to hover0) The power required to fly at various
air speeds has been esti-) mated (Pennycuick 1969; Greenwalt 1975),
and these estimations suggest that there exists for each bird,
depending on the aerodynamics of its flight, a speed at which the
rate of power output is minimal (Vmp; see Fig® 10)® Laboratory
measurements by Tucker (1968) have confirmed the general shape of
the predicted curve® To hover for the least cost, a bird should
select a wind speed equivalent to Vmp, and thus one might expect
hovering to be done most often when winds are of this velocity® But
wind speed varies with altitude® The exact pattern of this
relationship is complex and depends on several factors (see Lowry
1969) but, in general, wind speed increases with height above the
ground
-
Powe
r Re
quire
d to
Fly
Vmp
Air Speed
Figure 10. The effect of air speed on the power required to fly.
— (From Pennycuick 1969.) Vmp is the air speed at which flight cost
is minimized.
-
53surfaceo Thus, under a fairly wide range of conditions, a bird
should be able to find the optimal wind velocity by varying its
altitude„ But the existence of an optimal hunting height
(determined in large part by prey size and density) restricts the
range of altitudes where hovering is profitableo Optimal wind speed
and optimal hunting height will coincide only at certain times;
hovering when they do not coincide increases foraging costs or
decreases prey captures and thus lowers net energy intake
-
Prediction 3s Most hovering should be done at times when optimal
wind speed for hovering and optimal hunting height coincide=
Whereoptimal hunting height is nearly constant, most hovering would
be ex-
jpected at times of an optimal wind speed measured at a fixed
height with a decreasing amount of hovering at increasing
deviations above or below this optimumo
Test of Prediction 2s Hovering kestrels captured food at a rate
7=6 to llo2 times greater than birds hunting from percheso The
increased energetic costs of hovering cannot be calculated
precisely but appear to be of the same magnitude as the above
numbers0 Total weight of prey items taken from hovers and perches
was calculated by assigning approximate weights to each prey item
captured.(Table 11)0 Approximate rates for energetic gain were
obtained by dividing the sum of these weights by total time spent
either perch-hunting or hovering (energy per gram is approximately
equal for insects and vertebrates; Cummins 1967)o The calculated
rate of prey taken from hovers was 7o6 times that taken from
perches«, In only 70% of attacks from perches and 4-7% of attacks
from hovers was the outcome (capture or escape of prey) observedo
If the same kinds and proportions of prey were taken in these
attacks as in those of unknown outcome, the rate of weight capture
from hovers would be 1102 times that from percheso However, most
attacks where outcome was unknown were probably unsuccessful, or if
successful, the prey taken was very small and immediately
cottsumedo Therefore this value probably gives an approximate upper
limit for the ratioo
-
55Table 11„ Comparison of prey types and rates of prey capture
from
perches and hovers0
Prey TypeApproxo Wgto (g)
PerchesTotal
# Taken WgtoHovers
Total # Taken Wgt=
Sigmodon spo 60 4 240 2 120Baiomys taylori 8 1 8 0Perognathus
flavus 8 0 = 1 8lizards1 k 5 20 02Scaphiopus hammondi 14 1 14 0
—Rana pipiens 2 6 12 0grasshoppers 1 43 45 5 5beetles o5 5 2=5 2
1Igeo arthropods 1 22 22 7 7stoo arthropods o2 6 1=2 13 2=6
Total weight 562=7 143=6Seconds watched Rate of weight
capture
119416 6332
(g/s x 1000) 3o0 22=7^Small Sceloporus undulatus
-
56Mammals, especially Sigmodon, account for a large proportion
of
the total weight of prey taken but a small percentage of total
individual s= This may result in some inaccuracy in comparisons of
rates of biomass due to sampling error= But even discounting the
mammals taken, prey biomass was captured from hovers at a rate 2o6
to 3=7 times that captured from percheso
The cost of flying at Vmp for the American Kestrel has not been
measured* However, data for three species of birds flown in wind
tunnels indicated that metabolic rates while flying at Vmp were
approximately 6 times resting metabolism (Bernstein, Thomas and
Schmidt- Nielsen 1975)° Greenwalt (1975) suggested that wind tunnel
measurements exaggerate flight costso Additionally, some hovers may
not be as costly as flying at Vmp because under certain conditions
updrafts provide some lift and a hovering bird does not have to
expend as much energy to remain stationary* Extreme "hangs" where
no wingbeats occur are the equivalent of static soaring and can
occur only when the sinking speed of the bird matches the vertical
component of the wind (updraft) and the flight speed of the bird
matches the horizontal component* Although I have recorded hangs
for kestrels that lasted as long as 17 s, most were short (
-
2o The energy required to reach hunting height was greater while
hovering0 Mean hovering height was 12=5 m whereas mean perch height
was 7°3 fflo
3o Total- energy spent flying between hunting locations was
greater when hoveringo Hovering kestrels searched approximately 2=4
times as many areas before an attack was made than did perched
birds (3=4 area changes per attack and 104 area Changes per attack,
respectively) =,
The relative costs of these activities are not known and thus
the costs of hovering cannot be compared quantitatively with those
of perch-huntingo Nonetheless, the relative magnitude of the
increased costs and benefits of hovering seem approximately equal
as predictedo
Test of Prediction J>% Hovering effort, i0eo, time
hovering/time observed, as well as the number of hovering birds
increased with wind speed measured at 2 m to a peak at a wind speed
range of 18-24 km/h and then decreased (Figo 11)= Virtually no
hovering occurred at times of no windo The shape of this curve is
as expected, but even when wind conditions were in the optimal wind
speed range, hovering occurred less than 8% of the timeo Note that
male kestrels generally hovered more than femaleso
More evidence that hovering is wind dependent is provided by
other observations. In several cases, kestrels paused as if to
hover during flights between lines of utility poles, but then
continued on without hovering. On two occasions kestrels even left
utility wires and flew parallel to them before returning as
described earlier and made similar pauses without hovering. These
aborted hovers took place
-
58
d^% time males hovered9 % time females hovered# % time all
kestrels hovereda % of all kestrels watched
that hovered
12.0
10- - T 100
8 — - - 807.8OZccLU>I G - - LUs
6.6 67;|.6 .467
- -6 0
4 - - - - 4 0
2.9
2.0 --202 “" 19 V
.59.31
Km/h 0-8.1 9.7-16.1 17.7-24.2 25.8-32.2 >32.2MPH 0 -5 6 -1 0
11-15 16-20 > 20
WIND SPEED
Figure 11. Hovering effort as a function of wind speed. — The
points for wind speeds greater than 32.2 km/h are based on
observations of only 8 birds for 4l min on one day.Samples for
males at wind speeds above 24.2 km/h were too small for
analysis.
% BI
RDS
WAT
CHED
TH
AT
HO
VER
ED
-
59on days when wind speed was extremely variable« Apparently,
wind conditions were not right when these birds began to hover0
Hovering HeightAs previously discussed, hovering height is
likely determined
by both wind speed and optimal hunting height = The latter in
turn is influenced by prey density and size* If these stay
reasonably constant, hovering height will be determined primarily
by wind speed arid the following prediction can be made0
Prediction 4: Because wind speed increases with height,
hoveringheight and wind speed measured at a fixed height should be
inversely related® The exact relationship depends on the relation
of wind speed to height, which is influenced by many variables, but
above 2 m it is often close to linear (Geiger 1966)
-
602o Ground topography can influence the wind speed-height
rela
tionship = Kestrels hovering in series that crossed ravines
stayed at the same absolute height even though the recorded height
above ground changedo The birds were probably staying in the same
air speed layero
3o Hangs are not restricted by the same wind speed-height rela-
tionshipo As mentioned earlier, hangs can occur only when the
vertical and horizontal components of the wind are matched by the
sinking speed and flight speed of the bird, respectively* But a
bird's sinking speed is in part determined by its air speed (Tucker
and Parrottl970)0 For a given updraft there is essentially only one
air speed at which the sinking speed will match* This wind speed
could be found by varying altitude as for flapping hovers, but the
height at which conditions are right for hangs depends on the
magnitude of the updraft as well as the wind speed* If hangs were
used exclusively, no correlation between hovering height and wind
speed measured at 2 m would be expected* The extent of the
deviations from the wind speed-hovering height curve caused by
hangs is probably determined by the stability and predictability of
the conditions necessary for hangs* If these conditions are stable
and predictable, kestrels may seek them because costs would be
significantly lower than flapping hovers; but, if not, hangs may
occur only opportunistically during flapping hovers* In the latter
case, although the cost of hovering would still decrease, the
hovering height should be determined primarily by the wind
speed-height relationship*
4* Tradeoffs between the optimal wind speed and optimal hunting
height may result in significant deviations from optimal wind speed
if
-
61optimal hunting heighir becomes an overriding factor= Such a
situation might be expected when prey were very abundant. In this
case, search time (and thus hovering time) might be so low that it
would be more profitable to hover low at a greater cost than to
expend the energy and time to fly up to optimal hovering height. On
three occasions I observed kestrels making very low (5 m or less)
series of hovers at times.when the wind speed-holering height
regression would have predicted a hovering height of approximately
15 m. In all of these cases, hovers were very short in duration and
nearly all resulted in attacks (8%, 4l of 50) suggesting that prey
(apparently ants) were very abundant.
Another possible consequence of a tradeoff between optimal
hunting height and wind speed was suggested by an unpredicted
pattern of hovering. In series of hovers uninterrupted by prey
attacks, se- quential hovers were significantly (X = 50.3, p
-
62speed should be equal to Vmp0 No one has measured Vmp for the
American Kestrel but it can be estimated from body weight and wing
span using empirical or theoretical equations. Unfortunately, these
equations predict values ranging from 8,1 to 42=1 km/h (see Table
12), A flight speed of 39°9 km/h has been measured for a kestrel in
the field (Tucker and Schmidt-Koenig 1971) but the purpose of the
flight, and hence whether it was likely to be at Vmp, was
unknown.
Because wind speed increases with height, the wind speeds I
measured at 2 m were not those at which kestrels were hovering, C,
Richard Tracy (personal communication 1978) has informed me that
the actual wind speeds at which kestrels hovered can be robustly
estimatedfrom my wind speed measurements by using the equation:
u* (z - d)u(z) = k In ( zQ ) (6)
where u(z) is the wind velocity as a function of height, z; u*
is the shear velocity, which can be calculated from measured wind
speeds at a known height; k is the Karmen constant, ,41; d is "zero
plane displacement,” 063 h; zq is "roughness length,” ,13 h, and h
is the average vegetation height. For the wind speed range measured
at 2 m of 17o7 to 24,2 km/h, at which kestrels hovered most
frequently, hovering heights calculated from the linear regression
equation of the wind speed-height relationship range from 11,1 to
14,3 m. Wind speeds at these heights calculated from the above
equation are 24,8 to 32,5 km/h. This range of values compares
favorably with the range of the middle four values of Vmp from
Table 2 which suggests that kestrels do indeed hover at wind speeds
equivalent to Vmp,
-
63Table 12o Calculations of Vmp for the American Kestrelo^
Vmp (km/h) Equation Source
8*1=760 (y)^ pVs/ Pennycuick 1969, P= 530
33=8 14=6 mo2° Tucker 1973, p= 70742ol (835=3 x 10”6h +
15=73)m°169 Tucker 1974, p= 30631=9 from Table 22 Greenwalt 1975,
P= 3726o 2 e*wB Greenwalt 1975, P° 4424c42 e
-
64If wind velocity were the only factor that affected
hovering
height, an estimate of the actual wind speed at which kestrels
hovered could be obtained by extrapolating the slope of the linear
regression line in Figure 12 to a height of 2 m= However, because
optimal bunging height should also affect hovering height, hovering
birds should Compromise between the. height where wind speed
equaled Vmp and the optimal hunting height0 Such a compromise would
result in a shallower slope of the regression line than would be
expected from considerations of wind speed alone, and extrapolation
of the line to 2 m should give a wind speed value greater than Vmp0
Extrapolation of the regression line gives a wind speed of 4l08
km/h which, although it is quite close to the flight speed recorded
by Tucker and Schmidt-Koenig and the highest calculated value for
Vmp, is substantially higher than most of the calculated values for
Vmp, as expectedo
Other factors may also affect the optimal wind speed at which to
hoverq Hovering at a given wind speed may not be equivalent to
flying at that air speed because wind speed is often quite variable
while the bird is hovering whereas a bird in forward flight can
maintain a relatively constant speedo If wind speed is variable
while hovering, the best mean wind speed at which to hover is one
that minimizes energy output over the range of wind speeds
experienced^ From the shape of the power required-air speed curve
shown by Pennycuick (1969; she also Fig* 10) it can be seen that
deviations from Vmp at higher speeds are less costly than those at
lower speeds because of a steeper slope below Vmp0 Thus, for a bird
hovering in a variable wind.
-
65
65 f
X 30
• •» •» •»>Xw\
•> •J55* •* •»>•»»•» •».•»•> e>• • • emX»»
0Km/h 0 MPH 0
WIND SPEED AT 2M
Figure 12. Hovering height as a function of wind speed at 2 m.
—Hovers over prey apparently seen prior are not included. In
situations where wind speed changed during a hover, a median value
was used. Line si is the linear regression line of best fit and
line b is the best logarithmic fit.
-
66it might be advantageous to select an average wind speed
slightly greater than Vmp0
Hangs could also affect the optimal wind speed at which to
hover= Hangs are restricted only to a range of wind speeds where
gliding is possible and not just to a wind speed equal to Vmp0 The
extent of this effect is determined by the stability and
predictability of the conditions when hangs are possible, as
mentioned earlier. The direction of the effect is determined by the
relative position of Vmp to the range of air speeds where gliding
can occur0
Hovering TimeTimes considered here are for hovers from which no
prey were
attacked, i0e,, giving-up times. For 185 such hovers, times
ranged from 3 to 112 s with a mean of 18,3 s.
According to optimal foraging theory, giving-up time in a patch
should be inversely proportional to prey density (Pyke et al,
1977)= Although few data on prey abundance were taken, it appeared
that prey became more scarce as the season progressed from
September to April,Mean duration of hovers for each month
significantly increased (r = ,76, p
-
Test of Prediction 5? A significant positive correlation existed
between hovering time and the time of the last hover from which
prey was attacked (r = o 4 9 » n = 39? p < o 0 0 5 ) o Sample
size was limited largely because many attacks ended sequences of
hoveringo
Hovering time might also be affected by profitability; if so,I
would make the following prediction*
Prediction 6: Hovers made when optimal wind speed and
optimalhunting height nearly coincided should be longer than those
hovers made when optimal wind speed and optimal hunting height
differed considerably*
Test of Prediction 6s When wind speeds measured at 2 m were
between 17=7 and 32*2 km/h, which was when kestrels hovered most
and apparently when optimal wind speed and optimal hunting height
most nearly coincided, hovers were significantly (t = 1*93? p
-
68considered to be worth more effort than others= Situations
where prey
were seen but not attacked would provide one source of such
informa
tion*
3o If wind conditions were right for hangs, hovering time might
be lengthened because of lower cost*
Hovering as an Alternate Hunting Technique Results of this study
show that hovering by American Kestrels
occurs in predictable patterns that, in general, support ideas
of optimal foraging theory* Hovering appears to be an alternate
foraging technique that complements hunting from perches* But high
costs restrict hovering to times when prey availability and wind
conditions are favorable* For hovering to be most profitable,
optimal hunting height must approximate the height where hovering
is least costly, which is determined by wind speed*
Hovering by other perch-hunting birds also appears to be
influenced by wind conditions* Schnell (196?) observed that
Rough-legged Hawks (Buteo lagopus) spent more time in the air at
times of higher wind speed, and I have observed that this species
hovers at greater heights than kestrels at comparable wind speeds*
Rough-legs should hover at greater heights for two reasons: because
they have heavierwing-loading than kestrels, they should hover at
the least cost at higher wind speeds and thus at greater heights;
and because they take larger prey on the average, optimal hunting
height should be higher*I have also observed that the hovering of
Say*s Phoebes (Sayornis saya) and Mountain Bluebirds (Sialia
currucoides) is affected by winds and
-
69normally occurs at much lower heights than those of kestrels,
as would be expected for these smaller birds*
Some species that hover habitually, such as White-tailed Kites
(Elanus leucurus), Ospreys (Pandion haliatus), and some terns,
rarely hunt from perches, although they may hunt from flight
without hovering* Wind conditions should still affect hovering
flight in these species, and selection should favor the development
of aerodynamic properties that make hovering least costly at
prevailing winds* For White-tailed Kites wing-loading has been
reduced to a point where hovering is apparently profitable even
without wind (Balgooyen 1976)*
As 1 mentioned earlier, many species are capable of hovering;
some that normally do not hover may take advantage of this
technique when wind conditions are favorable* I have seen
Red-tailed Hawks (Buteo .jamaicensis), a species that normally
hunts from perches, hovering at times of high winds* However, most
were immature birds which probably have lighter wing-loading than
adults*
The use of hovering ranges from a rarely-used alternate hunting
technique for some species to one of the primary techniques for
others* Habitual hovering is restricted by environmental
conditions, morphology, and the suitability of other techniques*
Species that frequently hover normally occur in areas without
perches and many occur in areas of fairly predictable winds*
Hovering may be encouraged by competition with other species*I
have other data that suggest use of perches may be restricted by a
dominant species and hovering may enable a second species to
coexist*A similar situation may explain why male kestrels hover
more than
-
70females; females apparently force smaller males into areas of
lower perch qualityo Thus, hovering would be more profitable under
a wider range of conditionso Dominance may also explain my
observations that hovering is more frequent among immature
Red-tailed Hawks than adultsc
An animal’s foraging repertoire appears to be affected by prey
distributions, environmental conditions, and competition0
Repertoire size is in part determined by the predictability of
these factors and is limited by morphological restrictions,.
-
LIST OF REFERENCES s
Balgooyen, T« Go 1976» Behavior and ecology of the American
Kestrel (Falco sparverius) in the Sierra Nevada of California^UniVo
Califo Publo Zoolo 103:1-83*
Bent, Ao C* 1950o Life histories of North American wagtails,
shrikes, vireos, and their allies = UoSo Natlo Huso Bull* 197
=>
Bernstein,. Mo Ho, So Po Thomas, and K0 Schmidt-Nielsen* 1973°
Power input during flight of the Fish Crow (Corvus ossifragus)o Jo
Expo Biolo 38:401-4l0o
Cade, To Jo 1955o Experiments oh winter territoriality of the
American Kestrel, Falco sparveriuso Wilson Bull* 67:5-17o
Charnov, E» Lo 1973= Optimal foraging: - some theoretical
explorations* PhoD* Thesis, Univ* of Washington*
Collapy, Mo W* 1973= Predatory efficiency of American
Kestrelswintering in northeastern California* Raptor Res*
7:25-31*
Craig, Ro B* 1978= An analysis of the predatory behavior of the
Loggerhead Shrike = Auk 95:221-234*
Cummins, K= W * 1967= Caloric equivalents for studies in
ecological energetics* Unpublished manuscript*
Fox, R o , So W o Lehmkuhle, and D o H* Westendorf* 1976* Falcon
visual acuity* Science 192:263-265*
Geiger, R* 1966* The climate near the ground* Harvard Univ*
Press, Cambridge*
Greenwalt, C* H* 1975* The flight of birds* Trans* Amer*
Phil*Soc* 65:1-67* \
Grubb, T* C* 1977* Why ospreys hover* Wilson Bull*
89:149-150*Lowry, W * P* 1969* Weather and life* Academic Press,
New York*Mills, G* S* 1975* A winter population study of the
American Kestrel
in central Ohio* Wilson Bull* 87:241-247*Mills, G* So 1977*
American Kestrel rejects captured spadefoot toad*
Wilson Bull* 89:623*71
-
72Oaten, Ac 1977° Optimal foraging in patches? a case for
stochas-
tieity0 Theor° P o p e B i o l o 1 2 ? 2 6 3 - 2 8 5
oPennycuick, C0 J° 1969o The mechanics of bird migration* Ibis
111:525-556=Pinkowski, B* C* 1977= Foraging behavior of the
Eastern Bluebird*
Wilson Bull* 89:404-414*Pulliam, H* B* 1974* . On the theory of
optimal diets* Amer* Natur*
108:59-74=Pyke, G* H* 19780 Optimal foraging: movement.patterns
of bumblebees
between inflorescenceso Theor* Pop* Biol* 15:72-98*Pyke, G* Ho,
Ho Bo Pulliam, and E* Lc Char no v* 1977= Optimal foraging:
a selective review of theory and tests* Q* Bev* Biol*
52:I57-I54*
Boest, A* I* 1957= Notes on the American Sparrow Hawk* Auk
74:1-19=Schnell, G* D* 1967= Environmental influence on the
incidence of
flight in the American Bough-legged Hawk* Auk
84:175-182*Schoener, T* W* 1971= Theory of feeding strategies* Ann*
Bev* Ecol*
and Syst* 11:569-464*Sparrowe, B* D* 1972* Prey-catching
behavior in the Sparrow Hawk*
J* Wildlo Mgmt* 56:297-308*Tracy, C* Bichard* 1978= Personal
communication* Dept* of Zoology,
Colorado State Tlniv*, Ft* Collins*V* A* 19680 Bespiratory
exchange and evaporative water lossin the flying budgerigar* J*
Exp* Biol* 48:67-87=V* A* 1973= Bird metabolism during flight:
evaluation of atheory* J* Exp* Biol* 58;689-709=V* A* 1974* The
energetics of natural avian flight, pp* 298-333= In B* A* Paynter,
Jr* (ed*)* Ptibl* Nuttall Ornithol* Club No* 15=V* A* and G» C*
Parrott* 1970* Aerodynamics of gliding flight in a falcon and other
birds* J* Exp* Biol* 52:345-367=V* A® and K= Schmidt-Koenig* 1971=
Flight speeds of birds in relation to energetics and wind
direction* Auk 88:97-107®
Weis-Fogh, T* 1972= Energetics of hovering flight in
hummingbirdsand Drosophila* J* Exp* Biol* 56:79-104*
Tucker,
Tucker,
Tucker,
Tucker,
Tucker,
-
Weis-Fogh, T. 1973. Quick estimates of flight fitness in
hovering animals, including novel mechanisms for lift production#J.
Exp. Biol. 59:169-230#
Welty, J. C. 1962. The life of birds. W. B. Saunders Co.,
Philadelphia and London#