The Journal of Experimental Biology 200, 1527–1536 (1997) 1527

Printed in Great Britain © The Company of Biologists Limited 1997

JEB0864

HUMMINGBIRD HOVERING ENERGETICS DURING MOULT OF PRIMARY FLIGHT

FEATHERS
PENG CHAI*
Department of Zoology, University of Texas, Austin, TX 78712, USA

Accepted 10 March 1997

Summary
How does a hovering hummingbird compensate for the rates, wingbeat kinematics (stroke amplitude) and lift
loss of flight feathers during moult when the mechanism of coefficients did not vary during the moult. This constancy
lift force generation by flapping wings is impaired? The was mainly achieved through weight loss that alleviated
flight performance of five individual ruby-throated aerodynamic force requirements for weight support during
hummingbirds with moulting primary flight feathers and hovering. Since the metabolic power expenditure during
reduced wing area was compared with that before their moult was similar to that of normal birds but the
moult. Hummingbirds were flown in reduced air densities mechanical power requirement was reduced, the flight
using normoxic heliox so that a range of flight energetics efficiency also showed a sharp reduction during moult.
was displayed. The rate of moulting and the extent of wing This increased cost of flight may result from disruption of
area loss varied among individuals. One female could the integrity of the flight machinery. Overall, the control of
tolerate a 30 % loss of wing area in moulting and flew with body mass in hummingbirds can provide similar
only three outer primaries per wing. Further exploratory aerodynamic, muscle mechanical and physiological
study using the artificial reduction of wing area, either by capacities under conditions of variable flight demand.
cutting the tips of the outer primaries of a male or by
plucking the secondaries of two females, suggested that
secondaries play a minor role in lift force generation during Key words: air density, body mass, flight feathers, heliox, hovering
hovering whereas the tip area of primaries is crucial. For flight, hummingbird, Archilochus colubris, lift coefficient, moult,
the five birds, ranges of whole-bird oxygen consumption muscle power and efficiency, oxygen consumption, wing area loss.

Introduction
Moulting of the feathers must be a very energetically and is the first such study investigating moulting effects on the
stressful period in the annual cycle of birds. The costs flight energetics of birds.
associated with moulting are manifold given that flight Air density is a major determinant of aerodynamic power
performance, behavioural repertoire, plumage insulation and requirements (Norberg, 1990), and flight mechanics of birds
water repellency may all be impaired (Jenni and Winkler, can be conveniently varied by manipulation of the density of
1994). Furthermore, the efficiency of feather production is very normoxic gas mixtures (Dudley and Chai, 1996). Chai and
low, and moulting is energetically expensive and involves Dudley (1995) determined the limits of flight performance in
processes other than just the synthesis of keratin (Lindström et hummingbirds hovering in normoxic but hypodense mixtures
al. 1993). On average, energy expenditures required for of air and heliox. Ruby-throated hummingbirds demonstrated
moulting during the moulting period are 20–40 % of the basal considerable power reserves, and the maximum aerobic
metabolic rate (reviewed in Walsberg, 1983). capacity at low air densities was unequivocally indicated when
Moulting may be particularly challenging to hummingbirds aerodynamic failure caused the birds to descend to the chamber
because flight is essentially the only means of locomotion in floor after briefly hover-feeding. In this study, density
this taxon, and missing flight feathers will obviously impair the manipulation is used to elicit variable flight energetics of
mechanism of lift force generation by flapping wings. How do moulting ruby-throated hummingbirds and to identify the
hovering hummingbirds compensate biomechanically? How limits to hovering performance.
are the aerodynamic, mechanical and physiological parameters
of flight performance altered throughout a moult? This study
evaluates hovering performance of ruby-throated Materials and methods
hummingbirds during the moulting of primary flight feathers Ruby-throated hummingbirds (Archilochus colubris L.)

*e-mail: [email protected].
1528 P. CHAI
were mist-netted in the vicinity of Austin, Texas, during the primaries. This point occurred when the birds only had three
autumn migratory season of 1995 and were housed in screen or four old, outer primaries (the birds generally shed the first
cages of dimensions equal to those of the experimental cube. six or seven primaries over a short period). The other trial was
Hummingbird care was in accordance with federal and state arbitrarily chosen and was implemented either before or after
guidelines. Two or three birds were housed together within one the trial with smallest wing area. In this case, the old outer
cage. The experimental procedures and equipment have been primaries varied in number from two to four.
described previously (Chai and Dudley, 1995, 1996), and only Flight experiments were implemented within an airtight
a brief account will be given here. Birds were trained to feed acrylic cube (90 cm×90 cm×90 cm). For each trial, data were
through a cylindrical mask attached to a hanging syringe. Five collected initially from birds hover-feeding in unmanipulated
individual hummingbirds with initially undamaged flight air at pressures close to sea-level pressure. Air within the cube
feathers (one adult male bird 1M, one juvenile male bird 2M was then gradually replaced by filling with normoxic heliox
and three females birds 3F, 4F and 5F) were used in (79 % He and 21 % O2, density 0.40 kg m−3; the normobaric
experiments. Each bird was subjected to three trials. The first atmosphere exhibiting a density of 1.20 kg m−3) while allowing
served as the control and was conducted in the winter the cube contents to escape from an additional port. Hover-
(representing wintering, maintenance condition) prior to their feeding flight was then recorded approximately every
moult in the spring. Two trials were conducted during the 15–20 min when the bird was hungry and came to feed. The
moulting of primaries (Fig. 1). Since the most stressful point duration of hover-feeding flight t was timed from video
during moult was probably when the bird lost most of its recordings. Heliox filling was terminated after the bird
primaries to yield the smallest wing area and presumably demonstrated aerodynamic failure in hovering; reverse
smallest lift-generating capacity, one trial was timed at the pumping of ambient air was then initiated. The trial was
point when the bird was judged to possess the smallest area of generally stopped when the bird could then hover for increased
durations and feed normally.
Density reduction associated with replacement of normal air
A B by heliox was monitored acoustically (Dudley, 1995).
Metabolic power input during hovering was obtained from
10 10 .
measurements of rates of oxygen consumption (VO∑) using an
9 9
open-flow feeder-mask respirometry system. Oxygen
8 8 consumption was expressed at STP. Horizontal projections of
7 wingbeat kinematics of each hover-feeding sequence were
6 video-recorded from a mirror oriented above the bird at 45 ° to
5 6 the horizontal (filming and subsequent analysis were at
4 5 60 fields s−1 with a high-speed shutter of 1/4000 s). Wingbeat
4 kinematics recorded at each hover-feeding sequence, together
3
2 with morphological parameters for individual birds, were used
1
to estimate the mechanical power requirements of flight using
1
2 a detailed aerodynamic model of insect hovering flight
3 6 (Ellington, 1984a–f). Wingbeat kinematics measured for each
4 5
hovering sequence included wingbeat frequency n and stroke
3 cm amplitude Φ. Wingbeat frequency was derived from the
C interaction frequency between wing motion and filming rate,
and stroke amplitude was derived from video images in which
10 the wings were located at the extreme position of the wingbeat.
9 Morphological parameters used in aerodynamic calculations
included body mass m, relative wing mass mw for both wing
pairs and expressed as a fraction of body mass, wing length R,
total wing area S (the area of both wing pairs, body area not
included), wing loading pw (mg/S, where g is gravitational
acceleration) and aspect ratio AR (4R2/S). For humane reasons,
none of the experimental birds was killed for post mortem
Fig. 1. Configuration of the flight feathers of moulting ruby-throated
analysis. Wing mass and its spanwise distribution were
hummingbirds. (A) Unimpaired wing (bird 4F) showing primary and
secondary flight feathers; moulting of primaries (outward) precedes
estimated from one bird of each sex in previous studies with
that of secondaries (inward). (B) Only three old, outer primaries 8–10 the same species (Chai and Dudley, 1995, 1996). Wing mass
left, new primaries 4–6 growing, and all old secondaries plucked (bird and its distribution for moulting hummingbirds differ in an
4F). (C) More advanced moulting stage with two old, outer primaries unknown way from those of unimpaired wings. Possible
9–10 left, primaries 1–8 and secondaries 1–2 growing; the dashed line biomechanical implications of this difference for inertial power
indicates where the tips were cut (bird 2M). expenditure are treated in the Discussion.
 Hovering energetics in moulting hummingbirds 1529
.
For each hovering flight sequence, the mechanical power represent the values of the metabolic variables VO∑ and PIN near
requirements of flight were estimated by evaluating individual failure.
components of profile (P*pro), induced (P*ind) and inertial power In addition to studying the consequences of the normal
during the first half of a half-stroke (P*acc). Total power moult, an exploratory study was conducted by artificially
expenditure for a flight sequence was calculated for the two reducing the wing area and altering the wing shape. In order
cases of zero (P*zero) and perfect (P*per) elastic storage of wing to evaluate the role of primaries and secondaries in lift
inertial energy, which represent the maximum and minimum generation, the tips of the outer primaries (9–10) of the juvenile
estimates of required mechanical power, respectively male (bird 2M) were cut, and all the secondaries of two females
(Ellington, 1984f). Thus, P*zero=(P*pro+P*ind+P*acc)/2, assuming (birds 4F and 5F) were plucked (see Fig. 1, at this time, moult
zero elastic energy storage, and P*per=P*pro+P*ind, assuming of secondaries had not started). These manipulations were
perfect elastic energy storage. P*zero and P*per are expressed in performed after data for normal moulting had been collected.
muscle mass-specific form, assuming that flight muscle equals
25 % of the body mass (Chai and Dudley, 1995). However,
during moult, flight muscle mass may not remain a constant Results
fraction of body mass, and absolute muscle mass may not The moult pattern of captive ruby-throated hummingbirds
change. To evaluate this alternative assumption, a new value, follows the patterns described for wild individuals of this
P*per(fixed), was calculated assuming that muscle mass was species when they moult at their wintering ground (Baltosser,
fixed and did not change during moult, and that muscle mass 1995). The present study only concerns the moult of primaries
was 25 % of body mass before moult (see Table 1). Muscle that precedes moulting of secondaries (Fig. 1). Beginning with
mechanical efficiency ηm was estimated as P*per/(0.9P*IN), primary 1, moult proceeds outwards, with primaries being
assuming a 90 % direct energy expenditure by flight muscle replaced sequentially up to primary 8, after which primary 10
(Lasiewski, 1963; Wells, 1993). P*IN was derived from is shed before primary 9. Our captive birds tended to shed
measured rates of oxygen consumption and was expressed in primaries 1–6 or 1–7 over a relatively short period, and mainly
mass-specific form. P*per was used for comparative purposes relied on primaries 7–10 or 8–10 for flight while new primaries
because hummingbirds can probably store kinetic energy grew. The nonsequential moult of the outer primaries in
elastically during the deceleration phase of the wing stroke hummingbirds is unique among birds and suggests their
(Wells, 1993). In addition, because of body mass variation particular importance in flight (Epting, 1980). Individual
across the study period, whole-bird metabolic power input PIN variation with respect to the timing and rate of moult and the
.
(body mass × VO∑, unit changed to W) and whole-bird extent of wing area loss was also evident. Moulting was most
mechanical power output POUT (body mass × P*per in W) were impressive in bird 4F in that she quickly shed primaries 1–7
also calculated for comparison. Thus, ηm was also equal to (loss of 30 % of wing area) and flew using mainly the three
POUT/(0.9PIN). outer primaries (Fig. 1B).
The effects of normoxic density reduction and wing area loss Density reduction always led to aerodynamic failure
due to moult were evaluated for each of the ten kinematic, (Fig. 2A). The reduced wing area in moulting birds should
aerodynamic, mechanical and metabolic parameters (see cause failure at higher air densities than those of control birds
Tables 1, 2) using repeated-measures analysis of variance with unimpaired wings. However, the two males actually failed
(ANOVA) (SAS Institute, 1989). Both treatments (moult and at lower air densities during moult, whereas the three females,
density reduction) had three levels (see below), and these and as expected, exhibited failure at higher densities. This
their potential interactions were modelled as within-subject discrepancy reflects sexual dimorphism and the differences in
sources of variation. Because 30 tests (3×10) were conducted, patterns of wing loading between sexes; male ruby-throated
sequential Bonferroni adjustment for multiple statistical tests hummingbirds possess shorter wings and higher wing loading
was conducted to achieve a table-wide type I error rate of 0.05 (Fig. 2C). When moulting flight feathers, all individuals lost
(Holm, 1979). body mass despite the availability of food ad libitum (Fig. 2B),
Hover-feeding events were first grouped by rounding to the but the extent of this loss in males was such that their wing
nearest 0.1 kg m−3 the air densities at which feeding events loading was actually lower than that prior to moult, whereas
occurred. Flight performance recorded in normal air (the wing loading remained similar before and after moulting for
highest density level at 1.2 kg m−3) and at aerodynamic failure females. This result may explain the unexpected pattern of
(the lowest) were treated as separate density levels. Because aerodynamic failure in the two males with respect to density
the birds showed aerodynamic failure at different density levels reduction. In the moulting stages considered in the present
ranging from 0.5 to 0.9 kg m−3, a mean value across density study, the outer primaries (9–10, Fig. 1) were still intact, and
levels between normal air and aerodynamic failure was thus the wing length was unaffected. Aspect ratio then
calculated to achieve a balanced statistical design. Thus, three increased with decreasing wing area (Fig. 2D).
density levels were used for each trial. Because oxygen During moult, wingbeat frequency was slightly but
consumption rates at failure could not be reliably obtained significantly reduced (Fig. 3; Tables 1, 2). Stroke amplitude
given the short duration of hover-feeding, measurements at the did not change significantly, with aerodynamic failure
density level closest to aerodynamic failure were used to occurring at amplitudes of 170–180 ° (Fig. 3; Tables 1, 2).
1530 P. CHAI
Bird 1M 2M 3F 4F 5F 60
(40 mm) (43 mm) (49 mm) (48 mm) (49 mm) A
0.9 58
A
ρ (kg m−3)

0.8 56
0.7 54

n (Hz)
0.6 52
50
0.5
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 48
4.5 B 46
4.0 44
m (g)

3.5 42
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77
3.0 185
B
2.5 180
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 175
45 C 170

Φ (degrees)
pw (N m–2)

40 165
35 160
30 155
25 150
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 145
12
D
140
10 135
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77
AR

8 Remaining wing area (%)

6 Fig. 3. Wingbeat frequency n (A) and stroke amplitude Φ (B) in

C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 relation to wing area loss due to moult and air density reduction with
Remaining wing area (%) heliox (order of individuals as in Fig. 2). Solid lines link values at
aerodynamic failure; dashed lines link values in normal air.
Fig. 2. Air density ρ at aerodynamic failure (A) of five individual Intermediate values are means across density levels. Values in the
ruby-throated hummingbirds and their morphological indicators (B, boxes are from normal, unimpaired wings and serve as the control
body mass m; C, wing loading pw; and D, aspect ratio AR) in relation (symbol C on the x-axis).
to wing area loss due to moult. Values in the boxes are from normal,
unimpaired wings and serve as the control (symbol C on the x-axis).
Wing length in mm is indicated for each bird. (Fig. 5A), P*per with reduced . wing area showed a generally
downward trend (Fig. 5B). VO∑ showed a steep and significant
increase (Fig. 6A) with an associated and significant decrease
Thus, overall wing kinematics in terms of muscle contractile in muscle efficiency ηm (Fig. 6B; Tables 1, 2). During moult,
velocity (proportional to wingbeat frequency × stroke all birds showed reduced body masses. As a result, patterns of
amplitude) remained similar across moulting stages. During whole-bird metabolic and mechanical performance were
moult, the two aerodynamic variables (mean Reynolds number substantially different from their mass-specific values (Fig. 7;
— —
Re and mean lift coefficient CL ) did not show statistically Tables 1, 2). PIN showed no significant upward trend, whereas
significant changes (Fig. 4; Tables 1, 2). Wing area and mean POUT showed a steep decline with reduced wing area.
wing chord were reduced during moult, and this resulted in Statistical results for density effects on kinematic, mechanical
—
smaller values of Re . However, because females showed and metabolic variables were in close agreement with earlier
higher failure densities at aerodynamic failure, this increase studies on normoxic density reduction using heliox (Chai and
was a consequence of the decrease in wing area and resulted Dudley, 1995, 1996). Density reduction significantly affected
—
in an inconsistent distributional pattern of Re between sexes. all variables except for muscle mechanical efficiency; ηm was,
—
The individual P value of moult effect on Re is 0.006 however, significantly influenced by moult (Table 2).
(Table 2). However, after sequential Bonferroni adjustment,
this P value became marginally insignificant. The Artificial wing area and shape manipulation
—
distributional pattern of CL did not indicate any effect of moult. In the exploratory study (no statistics were performed owing
Depending on the extent of elastic energy storage, the effect of to the small number of samples), cutting off the tips of
moult on muscle mass-specific power output could be primaries 9 and 10 of bird 2M resulted in a smaller loss of wing
significant (Fig. 5; Tables 1, 2). While P*zero showed no trend area (10 %) than plucking out all the secondaries (17 % for bird
 Hovering energetics in moulting hummingbirds 1531
Table 1. Summary of morphological, kinematic, mechanical Table 2. Results of repeated-measures ANOVA for kinematic,
and metabolic variables for five birds in relation to moulting mechanical and metabolic variables assessing effects of wing
stages area reduction due to moult and air density reduction in
Moulting stages heliox

Moulting, Moulting, P value

Before moult, intermediate greatest Variables Moult Density Moult × Density
Variables wings unimpaired wing area loss wing area loss
n 0.001* (2,8) 0.001* (2,8) 0.927 (4,16)
ρfail (kg m−3) 0.64±0.13 0.65±0.06 0.73±0.08 Φ 0.058 (2,8) 0.001* (2,8) 0.780 (4,16)
—
m (g) 4.2±0.1 3.1±0.2 3.2±0.3 Re 0.006 (2,8) 0.001* (2,8) 0.001* (4,16)
—
pw (N m−2) 38.0±5.9 32.6±4.4 36.0±2.0 CL 0.746 (2,8) 0.003* (2,8) 0.021 (4,16)
P*zero 0.030 (2,8) 0.001* (2,8) 0.176 (4,16)
AR 7.5±0.5 8.8±0.3 9.7±0.7
P*per 0.001* (2,8) 0.001* (2,8) 0.242 (4,16)
n (Hz) 49±4 47±3 47±4 V̇O2 0.001* (2,8) 0.001* (2,8) 0.146 (4,16)
53±4 50±5 50±4 ηm for P*per 0.001* (2,8) 0.136 (2,8) 0.130 (4,16)
Φ (degrees) 156±15 153±16 149±6 PIN 0.269 (2,8) 0.001* (2,8) 0.466 (4,16)
177±5 178±2 173±5 POUT 0.001* (2,8) 0.001* (2,8) 0.002* (4,16)
—
Re 9100±500 7200±400 6500±400
5400±800 4600±500 4700±900 P values are from univariate F-test.
— d.f. (given in parentheses), degrees of freedom.
CL 1.38±0.12 1.38±0.17 1.49±0.11
1.75±0.10 1.62±0.26 1.61±0.23 * significant on the basis of sequential Bonferroni adjustment with
P*zero (W kg−1) 250±25 254±23 240±21 table-wide type-I error rate at 0.05.
—
343±18 373±22 350±25 n, wingbeat frequency; Φ, stroke amplitude; Re , mean Reynolds
—
number; CL, mean lift coefficient; P*zero and P*per, mechanical power
P*per (W kg−1) 101±4 93±2 95±3
output per unit flight muscle mass assuming zero and perfect elastic
131±6 119±2 117±4
energy storage; V̇O2, body mass-specific rate of oxygen consumption;
P*per(fixed) 101±4 69±3 72±6 ηm for P*per, muscle mechanical efficiency; PIN, whole-bird metabolic
131±6 88±2 88±6 power input; POUT, whole-bird mechanical power output assuming
V̇O2 (ml O2 g−1 h−1) 41±2 53±3 56±8 perfect elastic energy storage.
47±4 64±6 67±5
ηm for P*per (%) 11.6±0.3 8.3±0.4 8.2±0.9
12.7±1.0 8.6±1.1 8.1±0.3 indicated by a greatly increased air density at aerodynamic
PIN (W) 1.02±0.05 0.98±0.07 1.04±0.16 failure (Fig. 8, Table 3). The two females could tolerate total
1.16±0.11 1.18±0.16 1.25±0.11 loss of secondaries with only a slight increase of air density at
POUT (W) 0.106±0.004 0.073±0.005 0.075±0.008 aerodynamic failure. Bird 4F was already operating with a
0.138±0.011 0.093±0.005 0.093±0.008 much reduced wing area, only 79 % that of unimpaired wings,
because of her relatively fast rate of moulting, and after
The first value of kinematic, mechanical and metabolic variables plucking out all her secondaries, she was flying with 62 % of
refers to the value in normal air; the second to the value at pre-moult wing area. Despite this, air density at aerodynamic
aerodynamic failure (V̇O2 and PIN close to aerodynamic failure). failure remained similar to that occurring before feather
Values at muscle mass-specific power output, P*zero and P*per, were plucking (Fig. 8), although both failure densities were higher
calculated assuming that flight muscle equals 25 % of the body mass. than that measured before the moult (0.53 kg m−3, Fig. 2A).
P*per(fixed), recalculated P*per assuming that muscle mass is fixed
For all three birds, artificial reduction of wing area shortened
and does not change during moult and that muscle mass was 25 % of
body mass before moult.
the hover-feeding duration (Fig. 8). This reduction was most
All values are given as mean ± S.D., N=5. evident for hovering duration in normal air, which tended to
ρfail, air density at aerodynamic failure; m, body mass; pw, wing be the longest in duration. Air density reduction using heliox
loading; AR, aspect ratio; n, wingbeat frequency; Φ, stroke amplitude; also shortened hover-feeding duration and eventually led to
— —
Re, mean Reynolds number; CL, mean lift coefficient; P*zero and P*per, aerodynamic failure. Hover-feeding duration at aerodynamic
mechanical power output per unit flight muscle mass assuming zero failure was very short and probably represented transient burst
and perfect elastic energy storage; V̇O2, body mass-specific rate of performance (Chai and Dudley, 1996), whereas hover-feeding
oxygen consumption; ηm for P*per, muscle mechanical efficiency; PIN, in normal air represented sustainable flight. Thus, the duration
whole-bird metabolic power input; POUT, whole-bird mechanical of sustained flight of all three birds was adversely affected by
power output assuming perfect elastic energy storage. artificial wing area reduction, whereas only the transient
performance of bird 2M was greatly curtailed. After cutting the
4F and 18 % for 5F; see Fig. 1 and Table 3). However, the tips of the two outer primaries of bird 2M, the steep increase
reduction in wing length and wing tip velocity for bird 2M in mean lift coefficient at aerodynamic failure (to a maximum
caused by trimming the primaries exerted more drastic value of 2.35, Table 3) indicated that its wings had difficulty
aerodynamic consequences than the loss of secondaries, as generating sufficient lift force. This was probably the reason
1532 P. CHAI
10000 A 420 A
400
9000
380
8000 360

Pzero (W kg –1)
340
7000 320
Re

6000 300
* 280
5000 260
240
4000
220
3000 200
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 C 86 81 C 93 85 C 85 82 C 79 70 C 90 77

2.0 B 140 B
135
1.8
130
1.6 125
CL

Pper (W kg –1)
120
1.4 115
1.2 110
* 105
1.0 100
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77
95
Remaining wing area (%)
90
— —
Fig. 4. Mean Reynolds number Re (A) and mean lift coefficient CL 85
(B) in relation to moult and air density reduction (order of individuals C 86 81 C 93 85 C 85 82 C 79 70 C 90 77
as in Fig. 2). Solid lines link values at aerodynamic failure; dashed Remaining wing area (%)
lines link values in normal air. Intermediate values are means across
Fig. 5. Muscle mass-specific power output assuming zero (P*zero, A)
density levels. Values in the boxes are from normal, unimpaired wings
and perfect (P*per, B) elastic energy storage in relation to moult and
and serve as the control (symbol C on the x-axis).
air density reduction (order of individuals as in Fig. 2). P*zero and P*per
were calculated assuming that flight muscle equals 25 % of the body
mass. Solid lines link values at aerodynamic failure; dashed lines link
for aerodynamic failure at a much higher air density. However, values in normal air. Intermediate values are means across density
this bird compensated for its wing damage by changing its levels. Values in the boxes are from normal, unimpaired wings and
wingbeat kinematics after 8 days (both wingbeat frequency and serve as the control (symbol C on the x-axis).
stroke amplitude increased) to improve lift-generating
capacity, giving lift coefficients comparable to those before the
operation (Table 3). (During this period, bird 2M had shed increased wingbeat frequency (Pennycuick, 1990). However,
primary 10 but primary 8 grew longer, so that the total wing wingbeat frequency was slightly but significantly reduced (Fig.
area remained the same.) Hover-feeding duration in normal air 3). This reduction is difficult to explain, but may relate to
was lengthened, but air density at aerodynamic failure physiological changes in the thoracic musculature or increased
remained similar to that immediately following the operation variability in wingbeat kinematics associated with feather loss.
(Fig. 8, cf. After 1 and 2). Various physiological systems involved in neural control,
oxygen and fuel supply, muscle mechanics, wingbeat
kinematics, etc., must be efficiently linked for a bird to
Discussion function in flight. Each such system should be flexible enough
Wing mass and its distribution in moulting hummingbirds to respond to variable ecological and physiological demands
were not measured in the present study, and values from birds on flight, such as those occurring during migration, courtship
with unimpaired wings were used instead. Since no noticeable and moult (Rayner, 1988, 1995). In the present study, it has
change in muscle distribution along the wing was evident (P. been shown that ruby-throated hummingbirds maintained their
Chai, personal observation) and hummingbird wing mass is ability to hover given a wide range of body masses, air
primarily restricted to the wing base (Wells, 1990), wing mass densities, wing area and wing shape and also that muscle mass-
and its distribution in moulting hummingbirds were unlikely specific rates of oxygen consumption could vary substantially.
to show significant variation. Both wing area and moment of This finding suggests broad responsiveness in the flight motor.
inertia for the virtual mass of the wing pair were reduced in However, the ranges of whole-bird oxygen consumption rate,
moulting hummingbirds, and this should be correlated with wingbeat kinematics and lift coefficients remained similar,
 Hovering energetics in moulting hummingbirds 1533
Table 3. Artificial wing area reduction on three ruby-throated hummingbirds during moult and the effects on morphological,
kinematic, mechanical and metabolic variables
Tips of two outer primaries cut All secondaries plucked
Bird 2M Bird 4F Bird 5F
Variables Before After 1 After 2 Before After Before After
Wing area reduction (%) 5 15 15 21 38 10 28
ρfail (kg m−3) 0.71 0.87 0.88 0.65 0.66 0.57 0.60
m (g) 2.84 2.79 2.80 3.21 3.14 3.07 3.16
R (mm) 43 36 36 48 48 49 49
pw (N m−2) 31.9 35.0 35.5 31.2 38.7 27.9 35.9
AR 8.5 6.6 6.7 9.1 11.6 8.9 11.1
n (Hz) 48 50 53 46 45 44 45
51 51 56 48 49 46 48
Φ (degrees) 153 169 172 139 141 155 148
177 175 181 177 175 179 172
—
Re 6900 7000 7600 6900 5500 7900 6200
4900 5400 6200 4800 3800 4200 3600
—
CL 1.52 1.98 1.49 1.53 1.44 1.12 1.40
1.64 2.35 1.62 1.50 1.44 1.57 1.74
P*zero (W kg−1) 243 234 290 231 239 275 252
356 258 355 389 403 379 381
P*per (W kg−1) 90 96 100 92 96 90 90
111 109 114 119 126 118 118
V̇O2 (ml O2 g−1 h−1) 51 57 60 57 59 50 49
66 59 68 69 71 67 68
ηm for P*per (%) 8.3 8.0 7.9 7.8 7.8 8.5 8.7
7.8 8.1 7.7 7.2 7.5 8.1 7.8
PIN (W) 0.85 0.94 0.98 1.07 1.08 0.90 0.91
1.10 0.97 1.12 1.29 1.30 1.20 1.25
POUT (W) 0.064 0.067 0.070 0.074 0.075 0.069 0.071
0.079 0.076 0.080 0.096 0.099 0.090 0.093

Because of moult, wing area was already reduced relative to unimpaired wings (under Before) and was further reduced after the operation
(under After).
Kinematic, mechanical and metabolic variables show two values taken (first) in normal air and (second) at aerodynamic failure (V̇O2 and PIN
taken close to aerodynamic failure).
For each bird, all measurements were collected on the same day except for values under After 2 (bird 2M), which were collected 8 days later.
Wing area reduction indicates the percentage reduction relative to unimpaired wings before moult; ρfail, air density at aerodynamic failure;
—
m, body mass; R, wing length; pw, wing loading; AR, aspect ratio; n, wingbeat frequency; Φ, stroke amplitude; Re, mean Reynolds number;
—
CL, mean lift coefficient; P*zero and P*per, mechanical power output per unit flight muscle mass assuming zero and perfect elastic energy storage;
V̇O2, body mass-specific rate of oxygen consumption; ηm for P*per, muscle mechanical efficiency; PIN, whole-bird metabolic power input; POUT,
whole-bird mechanical power output assuming perfect elastic energy storage.

pointing to homeostasis and a set range of values to which an More studies are obviously needed to evaluate the role of body
individual hummingbird is adapted. Mass loss can alleviate mass control in hummingbirds during moult and other flight
aerodynamic force requirements for weight support, and, activities such as migration (Carpenter et al. 1993) and
therefore, by controlling body mass in relation to different courtship (Calder et al. 1990). Two other studies also showed
flight demands, hummingbirds can maintain a sufficient that moulting hummingbirds in captivity with reduced wing
aerodynamic performance. At the same time, physiological area typically lost mass (Wells, 1990; Hiebert, 1993), and this
parameters in relation to muscle contractile dynamics and also appears to be the case for other birds (e.g. King, 1968;
oxygen and fuel supply can be maintained without excessive Klaassen, 1995).
variation. However, mass loss presumably reduces fat reserves Dissimilar body masses prior to and during moult point to
and may jeopardize energy security. Thus, such a strategy may differences in body composition. In the present study, P*per and
be selected for only under extraordinary flight conditions. P*zero were calculated assuming that flight muscle equals 25 %
1534 P. CHAI
75 A 1.4 A
70
1.3
65
1.2
60
VO 2 (ml O2 g–1

PIN (W)
55 1.1
50
1.0
h–1)

. 45
0.9
40
35 0.8
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 C 86 81 C 93 85 C 85 82 C 79 70 C 90 77
0.16 B
14 B
13 0.14
12

POUT (W)
0.12
11
ηm(%

10 0.10
)

9
8 0.08

7
0.06
6 C 86 81 C 93 85 C 85 82 C 79 70 C 90 77
C 86 81 C 93 85 C 85 82 C 79 70 C 90 77 Remaining wing area (%)
Remaining wing area (%)
. Fig. 7. Whole-body metabolic power input PIN and whole-body
Fig. 6. Body mass-specific rate of oxygen consumption VO∑ (A) and mechanical power output POUT in relation to moult and air density
muscle mechanical efficiency ηm (B) assuming perfect elastic energy reduction (order of individuals as in Fig. 2). Solid lines link values at
storage in relation to moult and air density reduction (order of or close to aerodynamic failure; dashed lines link values in normal
individuals as in Fig. 2). Solid lines link values close to aerodynamic air. Intermediate values are means across density levels. Values in the
failure; dashed lines link values in normal air. Intermediate values are boxes are from normal, unimpaired wings and serve as the control
means across density levels. Values in the boxes are from normal, (symbol C on the x-axis).
unimpaired wings and serve as the control (symbol C on the x-axis).
diverse collection of insects, birds and bats. An explanation for
of body mass. However, it is not possible to determine these tight relationships is still elusive, but the reduced muscle
accurately the proportions of flight muscle and fat storage in mechanical efficiency (ηm) of moulting birds may indicate a
relation to varying body mass without immediate post mortem ‘penalty’ for straying from the laws of shape. To understand
analysis (ruby-throated hummingbirds can undergo rapid body the mechanisms and dynamics of aerodynamic force
mass change of up to 10 % in 1 day; P. Chai, unpublished generation of flapping wings more fully, further detailed
material). The values of P*per and P*per(fixed) represent the two investigation, using techniques such as high-speed video and
extremes (Table 1) in that P*per assumes that flight muscle mass vortex visualization to measure angle of attack, wing twist, and
remains a constant fraction of body mass whereas P*per(fixed) vortex formation and shedding, is needed.
assumes that flight muscle mass is invariant before and during Both whole-bird and mass-specific power output (POUT and
moult. The patterns of muscle and fat changes during moult P*per) declined during moult, whereas whole-bird power input
are presently unclear. (PIN) remained. stable and mass-specific rate of oxygen
The model developed by Ellington (1984f) uses consumption (VO∑) increased. Thus, hovering during moult
approximations to calculate profile drag coefficent (CD,pro); became more costly and muscle efficiency ηm was
thus, the value of P*pro is less accurate than the value of P*ind significantly reduced. To facilitate comparison, ηm was
(P*pro generally only accounts for one-quarter of P*per). Profile calculated assuming that 10 % of metabolic power was used by
drag and P*pro may be higher for a moulting bird with imperfect basal metabolism and maintenance needs, i.e. 90 % direct
wings. The wing shape of moulting hummingbirds also energy expenditure by flight muscles (Lasiewski, 1963; Wells,
deviates from Ellington’s (1984b) ‘laws of wing shape’, i.e. 1993). However, energy expenditures dedicated to moulting
strong positive correlations between ‘shape parameters’ of average 20–40 % of the basal metabolic rate (Walsberg, 1983).
spanwise wing area, mass and virtual mass distributions for a Even adjusting for this bias, the muscle mechanical efficiency
 Hovering energetics in moulting hummingbirds 1535
30

25 Bird 2M
Before
2
1 After 1
20 2 2 After 2

15 2
2 1
1
10 11
2
1 1 2
5 2
1
2
0
1.2 1.1 1.0 0.9 0.8 0.7 0.6
20
Bird 4F
Before
15 After

10

5
t (s)

0
1.2 1.1 1.0 0.9 0.8 0.7 0.6
40

35
Bird 5F
30 Before
After
25

Fig. 8. Hover-feeding duration t before and after 20

artificial wing area reduction and in relation to air
density reduction. The tips of the two outer 15
primary fight feathers of bird 2M were cut, and all
secondaries of the two females (birds 4F and 5F) 10
were plucked. Values in the boxes are from
normal air; circled values are at aerodynamic 5
failure. For each bird, all measurements (before
and after the operation; after 1 for bird 2M) were 0
from the same day except for values of After 2 1.2 1.1 1.0 0.9 0.8 0.7 0.6
(Normal air)
(bird 2M), which were collected 8 days later. Air density (kg m–3)

during moult was still greatly reduced (Fig. 6). Hummingbirds strain and rates of strain with minimal risk of damaging bones,
with damaged wings also augment their oxygen consumption muscles and tendons. Concurrently, the force balance achieved
and exhibit reduced flight efficiency (Epting, 1980; Wells, by the flapping wings should be stable. Impaired wings may
1990). increase the operating cost of flight muscles and their
Although ruby-throated hummingbirds demonstrate flexible regulation of flight stability through neural control.
physiological functions to cope with different flight demands, Artificial manipulation of flight feathers suggests that
the greatly reduced ηm during moult suggests that there is an secondaries play a minor role in lift force generation during
ideal flight design which links the aerodynamic properties of hovering, whereas the tip area of primaries is crucial. This
the wings, the contractile properties of the flight muscles and result is consistent with the hovering aerodynamic model of
the physiological features of oxygen and fuel supply. Moult Ellington (1984f). After cutting the tips of the two outer
disrupts the integrity of this flight machinery. For example, for primaries of bird 2M, mean lift coefficients rose sharply (see
a hummingbird to hover efficiently, wing morphology and After 1, Table 3). Such excessive lift requirements probably .
wingbeat kinematics should presumably operate in such a way contributed to early aerodynamic failure for this bird. The VO∑
that the supporting flight muscles can perform at their optimal of bird 2M close to aerodynamic failure at After 1 was lower
1536 P. CHAI
than that before the operation. This result suggests that oxygen ELLINGTON, C. P. (1984a). The aerodynamics of hovering insect flight.
supply was not limiting immediately after the operation. I. The quasi-steady analysis. Phil. Trans. R. Soc. Lond. B 305, 1–15.
However, 8 days later, at After 2, bird 2M was able to meet ELLINGTON, C. P. (1984b). The aerodynamics of hovering insect flight.
force requirements by altering wingbeat kinematics. The II. Morphological parameters. Phil. Trans. R. Soc. Lond. B 305,
duration of hover-feeding in normal air was restored to the 17–40.
ELLINGTON, C. P. (1984c). The aerodynamics of hovering insect flight.
level occurring before the operation, although air density . at III. Kinematics. Phil. Trans. R. Soc. Lond. B 305, 41–78.
aerodynamic failure did not change (Fig. 8). Since the VO∑ of ELLINGTON, C. P. (1984d). The aerodynamics of hovering insect flight.
bird 2M close to aerodynamic failure at After 2 was similar to IV. Aerodynamic mechanisms. Phil. Trans. R. Soc. Lond. B 305,
the level before the operation, oxygen supply may have been 79–113.
a limiting factor. Although bird 2M showed similar failure ELLINGTON, C. P. (1984e). The aerodynamics of hovering insect flight.
densities at After 1 and 2, two different types of constraints V. A vortex theory. Phil. Trans. R. Soc. Lond. B 305, 115–144.
might be responsible. Cutting the tips of primaries and ELLINGTON, C. P. (1984f). The aerodynamics of hovering insect flight.
plucking out secondaries obviously caused further disruption VI. Lift and power requirements. Phil. Trans. R. Soc. Lond. B 305,
to flight integrity, which was already degraded through 145–181.
moulting. Toleration of damage and the restoration of EPTING, R. J. (1980). Functional dependence of the power for hovering
performance present an interesting experimental context within on wing disc loading in hummingbirds. Physiol. Zool. 53, 347–352.
HIEBERT, S. (1993). Seasonal changes in body mass and use of torpor
which to evaluate flight capacity, and both natural and artificial
in a migratory hummingbird. Auk 110, 787–797.
disturbance of flight feathers can shed light on the relative roles HOLM, S. (1979). A simple sequential rejective multiple test
of aerodynamic, mechanical and physiological constraints. The procedure. Scand. J. Stat. 6, 65–70.
results of the exploratory manipulations discussed in this paper JENNI, L. AND WINKLER, R. (1994). Moult and Aging of European
point to complex physical and biological factors influencing Passerines. London: Academic Press.
both sustainable and transient capacities to generate lift force KING, J. R. (1968). Cycles of fat deposition and molt in White-
by flapping wings. crowned Sparrows in constant environmental conditions. Comp.
Biochem. Physiol. 24, 827–837.
I would like to thank R. Dudley, J. L. Larimer, and two KLAASSEN, M. (1995). Moult and basal metabolic costs in males of
anonymous reviewers for providing helpful comments on the two subspecies of stonechats: the European Saxicola torquata
rubicula and the east African S. t. axillaris. Oecologia 104,
manuscript and K. L. Schlegel for artwork. This work was
424–432.
supported by an NIH NRSA. LASIEWSKI, R. C. (1963). Oxygen consumption of torpid, resting,
active and flying hummingbirds. Physiol. Zool. 36, 122–140.
LINDSTRÖM, Å., VISSER, G. H. AND DAAN, S. (1993). The energetic
References cost of feather synthesis is proportional to basal metabolic rate.
BALTOSSER, W. H. (1995). Annual molt in ruby-throated and black- Physiol. Zool. 66, 490–510.
chinned hummingbirds. Condor 97, 484–491. NORBERG, U. M. (1990). Vertebrate Flight: Mechanics, Physiology,
CALDER, W. A., CALDER, L. L. AND FRAIZER, T. D. (1990). The Morphology, Ecology and Evolution. Berlin: Springer-Verlag.
hummingbird’s restraint: a natural model for weight control. PENNYCUICK, C. J. (1990). Predicting wingbeat frequency and
Experientia 46, 999–1002. wavelength of birds. J. exp. Biol. 150, 171–185.
CARPENTER, F. L., HIXON, M. A., BEUCHAT, C. A., RUSSELL, R. W. RAYNER, J. M. V. (1988). Form and function in avian flight. In Current
AND PATON, D. C. (1993). Biphasic mass gain in migrant Ornithology, vol. 5 (ed. R. F. Johnston), pp. 1–66. New York:
hummingbirds: body composition changes, torpor and ecological Plenum Press.
significance. Ecology 74, 1173–1182. RAYNER, J. M. V. (1995). Flight mechanics and constraints on flight
CHAI, P. AND DUDLEY, R. (1995). Limits to vertebrate locomotor performance. Israel J. Zool. 41, 321–342.
energetics suggested by hummingbirds hovering in heliox. Nature SAS INSTITUTE (1989). SAS/STAT User’s Guide, version 6, 4th edn.
377, 722–725. Cary, NC: SAS Institute Inc.
CHAI, P. AND DUDLEY, R. (1996). Limits to flight energetics of WALSBERG, G. E. (1983). Avian ecological energetics. In Avian
hummingbirds hovering in hypodense and hypoxic gas mixtures. J. Biology, vol. 7 (ed. D. S. Farner, J. R. King and K. C. Parkes), pp.
exp. Biol. 199, 2285–2295. 161–220. New York: Academic Press.
DUDLEY, R. (1995). Extraordinary flight performance of orchid bees WELLS, D. J. (1990). Hummingbird flight physiology: muscle
(Apidae: Euglossini) hovering in heliox (80 % He/20 % O2). J. exp. performance and ecological constraints. PhD thesis, University of
Biol. 198, 1065–1070. Wyoming, Laramie.
DUDLEY, R. AND CHAI, P. (1996). Animal flight mechanics in WELLS, D. J. (1993). Muscle performance in hovering hummingbirds.
physically variable gas mixtures. J. exp. Biol. 199, 1881–1885. J. exp. Biol. 178, 39–57.