See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/6194023

Flight Speeds among Bird Species: Allometric and Phylogenetic Effects

Article in PLOS Biology · September 2007

DOI: 10.1371/journal.pbio.0050197 · Source: PubMed

CITATIONS READS

283 2,986

5 authors, including:

Johan Bäckman Per G.P. Ericson

Lund University Swedish Museum of Natural History
62 PUBLICATIONS 2,341 CITATIONS 285 PUBLICATIONS 10,184 CITATIONS

SEE PROFILE SEE PROFILE

Olof Hellgren
Lund University
104 PUBLICATIONS 6,074 CITATIONS

SEE PROFILE

All content following this page was uploaded by Johan Bäckman on 26 May 2014.

The user has requested enhancement of the downloaded file.

 PLoS BIOLOGY

Flight Speeds among Bird Species:

Allometric and Phylogenetic Effects
Thomas Alerstam1*, Mikael Rosén1, Johan Bäckman1, Per G. P. Ericson2, Olof Hellgren1
1 Department of Animal Ecology, Lund University, Lund, Sweden, 2 Department of Vertebrate Zoology, Swedish Museum of Natural History, Stockholm, Sweden,

Flight speed is expected to increase with mass and wing loading among flying animals and aircraft for fundamental
aerodynamic reasons. Assuming geometrical and dynamical similarity, cruising flight speed is predicted to vary as
(body mass)1/6 and (wing loading)1/2 among bird species. To test these scaling rules and the general importance of
mass and wing loading for bird flight speeds, we used tracking radar to measure flapping flight speeds of individuals
or flocks of migrating birds visually identified to species as well as their altitude and winds at the altitudes where the
birds were flying. Equivalent airspeeds (airspeeds corrected to sea level air density, Ue) of 138 species, ranging 0.01–10
kg in mass, were analysed in relation to biometry and phylogeny. Scaling exponents in relation to mass and wing
loading were significantly smaller than predicted (about 0.12 and 0.32, respectively, with similar results for analyses
based on species and independent phylogenetic contrasts). These low scaling exponents may be the result of
evolutionary restrictions on bird flight-speed range, counteracting too slow flight speeds among species with low wing
loading and too fast speeds among species with high wing loading. This compression of speed range is partly attained
through geometric differences, with aspect ratio showing a positive relationship with body mass and wing loading, but
additional factors are required to fully explain the small scaling exponent of Ue in relation to wing loading.
Furthermore, mass and wing loading accounted for only a limited proportion of the variation in Ue. Phylogeny was a
powerful factor, in combination with wing loading, to account for the variation in Ue. These results demonstrate that
functional flight adaptations and constraints associated with different evolutionary lineages have an important
influence on cruising flapping flight speed that goes beyond the general aerodynamic scaling effects of mass and wing
loading.
Citation: Alerstam T, Rosén M, Bäckman J, Ericson PGP, Hellgren O (2007) Flight speeds among bird species: Allometric and phylogenetic effects. PLoS Biol 5(8): e197. doi:10.
1371/journal.pbio.0050197

bird flight speeds may deviate from the aerodynamic scaling

Introduction
rules. Flight adaptations related to the birds’ ecology and
According to fundamental aerodynamics the lift force (L) phylogeny may have consequences for their cruising flight
generated on a wing is related to flight speed (U) as: speeds, and different flight modes (continuous or intermit-
1 tent flapping) may constrain the birds’ speeds [2,10].
L ¼  q  CL  S  U 2 ð1Þ
2 A full evaluation of the applicability of aerodynamic
scaling rules must be based, not on theoretically derived
where q is air density, S is wing area, and CL is the lift
speeds, but on empirical measurements of airspeeds of a wide
coefficient [1–3]. In horizontal cruising flight L balances the
variety of bird species in natural cruising flight. Here, we
weight (m 3 g), and aircraft as well as animals are expected to
fly at or near a value of CL giving the maximum efficient lift- present tracking radar measurements of flight speeds of 138
drag ratio. Provided that this value of CL is about equal species from six main monophyletic groups [12], which were
among bird species (as required for dynamical similarity) [1], analysed in relation to biometry (m, S, and wingspan b) and
it follows that cruising flight speed among bird species is evolutionary origin (as reflected by phylogenetic group). All
expected to scale with body mass and wing loading (Q ¼ m 3 g/ speeds reported here refer to flapping flight at cruising
S) as U } m1/6 and U } Q1/2, respectively (with the former speeds of birds on migration. By restricting the analysis to
proportionality based also on the assumption of geometrical migration flight we expect the birds to fly at an airspeed close
similarity; i.e., S varies with m2/3). These scaling rules have also to that associated with maximum lift-drag ratio [13]. All
been used to compare general speeds of a wide range of speeds designate equivalent airspeeds (Ue) corrected to sea
flyers, from the smallest insects to the largest aircraft [1,4–6]. level air density [14,15].
In the absence of reliable measurements of the airspeed of
different bird species in long-distance cruising (migration) Academic Editor: Ben Sheldon, University of Oxford, United Kingdom
flight, theoretically derived flight speeds for species of
Received February 2, 2007; Accepted May 16, 2007; Published July 17, 2007
different mass and wing morphology have been used to
explore these scaling rules [4,5,7–10]. Deviations from the Copyright: Ó 2007 Alerstam et al. This is an open-access article distributed under
the terms of the Creative Commons Attribution License, which permits unrestricted
expected scaling exponent in relation to mass have been use, distribution, and reproduction in any medium, provided the original author
found because of departures from geometrical similarity— and source are credited.
larger birds often tend to have proportionately larger wing Abbreviations: AIC, Akaike information criterion; df, degrees of freedom
area and span [2,5,9–11]. There are additional possible * To whom correspondence should be addressed. E-mail: Thomas.Alerstam@ekol.
reasons, besides departure from geometrical similarity, why lu.se

PLoS Biology | www.plosbiology.org 1656 August 2007 | Volume 5 | Issue 8 | e197

 Bird Flight Speeds

Author Summary degrees of freedom (df) ¼ 25, and p , 0.0001), and falcons/
crows/songbirds showed a scaling exponent of 0.08 that was
Analysing the variation in flight speed among bird species is clearly smaller than expected (t ¼ 6.01, df ¼ 37, and p ,
important in understanding flight. We tested if the cruising speed of 0.0001). For the other four groups, the scaling exponents
different migrating bird species in flapping flight scales with body ranged between 0.12 and 0.20 and were not significantly
mass and wing loading according to predictions from aerodynamic different from the predicted value (p . 0.2). The correspond-
theory and to what extent phylogeny provides an additional ing scaling exponents of Ue in relation to Q differed
explanation for variation in speed. Flight speeds were measured
significantly from the predicted value of 0.5 among three of
by tracking radar for bird species ranging in size from 0.01 kg (small
passerines) to 10 kg (swans). Equivalent airspeeds of 138 species the groups, flamingo/pigeons/swifts (exponent 0.28, t ¼ 3.22, df
ranged between 8 and 23 m/s and did not scale as steeply in ¼ 5, and p ¼ 0.023), divers/cormorants/pelican/herons/storks/
relation to mass and wing loading as predicted. This suggests that crane (exponent 0.36, t ¼ 2.59, df ¼ 15, and p ¼ 0.021), and
there are evolutionary restrictions to the range of flight speeds that falcons/crows/songbirds (exponent 0.28, t ¼ 4.88, df ¼ 37, and p
birds obtain, which counteract too slow and too fast speeds among , 0.0001). For the remaining three groups, the scaling
bird species with low and high wing loading, respectively. In exponents ranged between 0.42 and 0.54 and were not
addition to the effects of body size and wing morphology on flight significantly different from the predicted value (p . 0.4).
speed, we also show that phylogeny accounted for an important To determine if there were geometrical differences in wing
part of the remaining speed variation between species. Differences shape associated with differences in mass and wing loading,
in flight apparatus and behaviour among species of different
we investigated whether or not aspect ratio scaled signifi-
evolutionary origin, and with different ecology and flight styles, are
likely to influence cruising flight performance in important ways. cantly with m and Q. Aspect ratio is a dimensionless measure
of wing shape (¼b2/S). We found significant departures from
isometry with aspect ratio scaling positively to m as well as Q
(p , 0.01 on the basis of all species [n ¼ 129] and p , 0.05 on
Results the basis of independent phylogenetic contrasts [n ¼ 17], for
Relationships between Ue and m and Q for all different both scaling relationships).
species are plotted in Figure 1, with the lines showing the We also investigated the explanatory power of m, Q, aspect
allometric relations according to reduced major axis regres- ratio, and phylogenetic group to account for the variation in
sions (Table 1). Mean airspeeds among the 138 species ranged Ue (Figure 2). Mass accounted for only a small fraction of the
between 8 and 23 m/s. Birds of prey, songbirds, swifts, gulls, variation in flight speed while, as expected, speed was much
terns, and herons had flight speeds in the lower part of this more closely correlated with wing loading. There was a
range, while pigeons, some of the waders, divers, swans, geese, significant positive correlation between Ue and aspect ratio,
and ducks were fast flyers in the range 15–20 m/s. but aspect ratio provided no improvement of general linear
Cormorants, cranes, and skuas were among the species flying models (based on Akaike information criterion [AIC] [17])
at intermediary speeds, about 15 m/s. The diving ducks when combined with Q or phylogenetic group.
reached the fastest mean speeds in our sample, with several A most potent factor to account for the variation in Ue was
species exceeding 20 m/s, up to 23 m/s (Protocol S1). phylogenetic group; species of the same group tended to fly at
The scaling analyses at the species level are robust against similar characteristic speeds. The groups including birds of
possible biases from few tracks per species and from within- prey and herons had on average slow flight speeds for their
species variation in speed (see Materials and Methods and mass and wing loading, while the average speed for groups
including songbirds and shorebirds fell above the overall
Table 1). Because species do not represent an evolutionary
scaling lines (Figure 1). Main phylogenetic group alone
independent data point, we also calculated scaling exponents
accounted for a substantial proportion of the variation in
by analysis of independent phylogenetic contrasts [16]
Ue (adjusted R2 ¼ 0.55), and a general linear model including
according to the procedure and phylogeny [12] presented in
both Q and phylogenetic group was the most satisfactory
Protocol S2. We used the well-resolved molecular phylogeny
model according to AIC (with adjusted R2 ¼ 0.64; Figure 2).
by Ericson et al. [12] for our phylogenetic analyses and
Our estimates of the explanation provided by the phylo-
classifications. The scaling results corrected for phylogenetic
genetic component, according to Figure 2, are likely to be
dependence agreed very closely with the exponents calcu-
conservative because of the broad grouping across the entire
lated on the species level (Table 1), demonstrating that the
modern bird phylogeny. If tighter monophyletic groups at the
scaling exponents for Ue in relation to m as well as Q (0.12 and
family level were used (20 phylogenetic groups), phylogenetic
0.32, respectively; phylogenetic contrast analysis) were small- group accounted for a fraction as high as 0.68 (adjusted R2;
er than the predicted values of 0.17 and 0.50, respectively. For F19,118 ¼ 16.4, and p , 0.001) of the variation in Ue, and for a
the scaling of Ue versus m, the difference from the predicted model including both phylogenetic group and Q this fraction
value was at the significance level of 0.05 for the phylogenetic increased to 0.71 (adjusted R2; F20,108 ¼ 16.4, and p , 0.001).
contrasts analysis, and the difference was not statistically However, these models had positive DAIC-values (þ8.1 and
significant for the sample of speeds adjusted for within- þ28.8, respectively) in relation to the best model in Figure 2
species variation (Table 1). and were thus less satisfactory when considering fit and
Within the different main phylogenetic groups (species complexity in combination [17].
level) as defined in Protocol S1 (see Figure 1), the scaling
exponents of Ue in relation to m were significantly smaller
Discussion
than the predicted value of 0.17 among two of the groups.
Swans/geese/ducks showed a remarkable negative scaling Two main results emerged from our analyses; (1) that flight
exponent of 0.15 (difference from prediction t ¼ 13.40, speeds among bird species scaled significantly differently with

PLoS Biology | www.plosbiology.org 1657 August 2007 | Volume 5 | Issue 8 | e197

 Bird Flight Speeds

Figure 1. Bird Flight Speeds (Ue; m/s) Plotted in Relation to Body Mass (kg) and Wing Loading (N/m2) for 138 Species of Six Main Monophyletic Groups
The lines show the scaling relationships Ue ¼ 15.9 3 (mass)0.13 and Ue ¼ 4.3 3 (wing loading)0.31 as calculated by reduced major axis regression for all
species (Table 1). All axes are in logarithmic scale. Inserts show means (6 standard deviations) for the six main phylogenetic groups in relation to these
scaling lines. Species of the same group tend to fly at similar speeds, and phylogenetic group is an important factor to account for the variation in Ue.
doi:10.1371/journal.pbio.0050197.g001

mass and wing loading than predicted from basic aerody- explanation for this result. The negative scaling exponent
namic principles and (2) that phylogenetic group contributed of Ue in relation to m for the swans, geese, and ducks may be
in a highly significant way to explain the considerable an effect of a reduced flight power margin with increasing
variation in bird flight speeds that remained, even after the size restricting the largest flyers like swans to fly close to the
biometrical dimensions of the bird species had been taken minimum power speed rather than at the faster speed
into account. associated with maximum effective lift-drag ratio [18,19].
Such constrained flight speeds for the largest flyers will also
Scaling of Flight Speed
have the effect of reducing the overall scaling exponents, thus
The scaling exponents fell below predicted values for both
providing another contributory explanation for the observed
of the tested relationships, for Ue versus m as well as Ue versus
results in this study.
Q. Predicted scaling exponents were based on the assump-
tions of geometrical and dynamical similarity. Could devia- Dynamical similarity is reflected by Reynolds number,
tions from one or both of these assumptions explain our which will differ between bird species in proportion to their
results? Earlier studies have demonstrated that bird species size (length dimension) and speed [20]. Reynolds number
are not, on average, geometrically identical, but larger species shows a 15-fold range among the species in our sample
tend to have proportionately longer wingspans and larger (ranging from approximately 25,000 to 375,000 based on
aspect ratios [2,5,10]. This was confirmed for the sample in mean wing chord, S/b, as length measurement). Such a range
the present study with aspect ratio scaling significantly of Reynolds number may well be large enough to give rise to
positively to m as well as Q. significant departures from dynamical similarity. The main
An overall scaling exponent of 0.14 for flight speed versus expected consequence would be a reduced coefficient of
body mass was calculated for theoretical flight speeds after frictional drag for birds with large Reynolds number (i.e.,
taking the slight positive allometry in wing size into account large and fast birds) leading to an increased optimal cruising
for a large sample of bird species [9]. This fits well with the speed among these species [14,20]. Thus, such a departure
corresponding exponent for observed speeds in this study, from dynamical similarity is expected to show up as an
making departure from geometrical similarity a likely augmented scaling exponent for Ue versus m (and also for Ue

PLoS Biology | www.plosbiology.org 1658 August 2007 | Volume 5 | Issue 8 | e197

 Bird Flight Speeds

Table 1. Allometric Relationships between Bird Flight Speed (Ue; m/s) and Body Mass (kg) and between Ue and Wing Loading (N/m2)

Relationship Sample n a 95% CI for a c 95% CI for c t p

Ue ¼ a 3 (mass)c All species 138 15.9 15.2–16.7 0.13 0.11–0.15 3.23 0.0015
All species with ntracks  10 56 16.0 15.0–17.0 0.13 0.10–0.16 2.74 0.0083
All species with ntracks  10, adjusteda 39 16.0 14.9–17.1 0.13 0.10–0.18 1.67 0.102
Phylogenetic contrasts 17 —b —b 0.12 0.07–0.16 2.09 0.052
Ue ¼ a 3 (wing loading)c All species 129 4.3 4.0–4.6 0.31 0.27–0.35 9.73 ,0.0001
All species with ntracks  10 55 4.8 4.4–5.2 0.28 0.24–0.32 9.32 ,0.0001
All species with ntracks  10, adjusteda 38 4.7 4.2–5.3 0.28 0.23–0.34 6.48 ,0.0001
Phylogenetic contrasts 17 —b —b 0.32 0.24–0.40 4.41 0.0004

Scaling relationships have been calculated by reduced major axis regression for logarithmic values of Ue, mass, and wing loading. Confidence intervals (CI) for the scaling coefficient (a)
and exponent (c) were calculated by bootstrapping (100,000 replicates) [33,34]. Test statistics for the difference between observed and predicted values of c are given by t  (degrees of
freedom ¼ n  1) and p-values. Predicted c for Ue versus mass and wing loading are one-sixth and one-half, respectively. The basis and procedure of the phylogentic contrast analysis are
presented in Protocol S2.
a
Ue adjusted for within-species variation in relation to vertical speed, tail- and cross wind components, and flock size.
b
Reduced major axis regressions for phylogenetic independent contrasts are calculated through origo, providing estimates of only the slope corrected for phylogentic dependence
(Protocol S2).
c
Scaling exponent.
doi:10.1371/journal.pbio.0050197.t001

versus Q), rather than a scaling exponent lower than expected for Ue versus Q may be the result of general evolutionary
as in this analysis. forces acting to increase cruising speeds for species with the
In view of the opposite effects on scaling exponents of lowest wing loadings and reduce speeds for species with the
departures from geometrical and dynamical similarity, highest wing loadings. The bird species in our analysis show
respectively [1], we conclude that only the departure from approximately a 10-fold difference in their range of Q (from
geometric similarity can explain why the scaling exponent for about 15 to 150 N/m2, Figure 1). With an observed scaling
Ue versus m falls significantly below one-sixth among birds in exponent for flight speed of 0.31, this range of Q is associated
cruising migratory flight. with a 2-fold (100.31 ¼ 2.0) difference in flight speed. However,
Do geometrical differences provide a sufficient explanation with a predicted scaling exponent of 0.5 we would have
also for the fact that the scaling exponent for Ue versus Q fell expected more than a 3-fold difference in cruising speed
clearly below the expected value of one-half? One way to (100.5 ¼ 3.2). Given that birds with low Q (about 15 N/m2) fly at
evaluate this is to calculate the scaling exponent for flight a speed about 10 m/s (as observed), species with high Q (about
speed versus span loading (m 3 g/b2, where b is wingspan). 150 N/m2) would fly at 32 m/s according to the general
Span loading is equivalent to wing loading divided by the aerodynamic scaling rules. This may well be impracticably
aspect ratio, and for birds differing in their geometric wing fast and difficult to reconcile with flight performance in
shapes cruising flight speed is expected to scale most closely situations of start, landing, flock manoeuvres, etc. Conversely,
with the square root of span loading (under geometrical given that birds with high Q fly at a speed about 20 m/s (as
similarity flight speed is predicted to scale with the same observed), species with low Q would fly at only about 6 m/s
exponent of one-half versus both span loading and wing according to the general aerodynamic scaling rules. Such very
loading) [5]. slow speeds will be disadvantageous because of sensitivity to
The scaling exponent for Ue versus span loading (species wind, vulnerability to predation, etc. Hence, it seems
level, exponent 0.36 with 95% confidence interval 0.31–0.40, reasonable to expect that there are evolutionary forces
n ¼ 129 and phylogenetic contrasts, exponent 0.37 with 95% operating to compress the range of cruising flight speeds
confidence interval 0.26–0.48, n ¼ 17) exceeded that versus Q among bird species [5] and thus reducing the scaling
(with corresponding exponents of 0.31 and 0.32, respectively, exponent for Ue versus Q. This compression of the range of
Table 1) although still falling significantly below the predicted flight speeds is attained partly through general geometrical
value of one-half. This suggests that the geometrical differ- differences between species (larger aspects ratios among
ences explain part, but not all, of the discrepancy between species with larger mass and wing loading, as discussed
observed and expected scaling of Ue versus Q. Departure from above), but additional unknown mechanisms, perhaps asso-
dynamical similarity will, in its most simple form (as reflected ciated with different kinematics of flight or different muscle
by differences in Reynolds number), contribute to an operation between species, seem to be required to fully
augmented rather than reduced scaling exponent in relation explain the restricted range of flight speeds among bird
to that predicted and can therefore not provide any useful species.
additional explanation in this case (see above). Still, dynam- Bounding flight seems to be a mode for small birds (mainly
ical differences of other kinds may exist for reasons that are passerines) to mitigate the costs of fast flight [1,2,10,21], while
notoriously difficult to predict for flapping flight. Future flap-gliding, used by many raptors, is associated with a
studies of vortex patterns associated with flapping flight of reduction in cruising flight speed [21]. Both of these styles
different species will be important to demonstrate possible of intermittent flight are used by species with low or
dynamical differences between species (see below). intermediate Q (Figure 1), and, having opposite effects on
We suggest that the unexpectedly small scaling exponent flight speed, they are unlikely to provide a sufficient

PLoS Biology | www.plosbiology.org 1659 August 2007 | Volume 5 | Issue 8 | e197

 Bird Flight Speeds

modes of flapping flight; (1) continuous flapping (e.g.,

shorebirds and ducks), (2) intermittent flapping with short
gliding phases (raptors, swifts, and swallows), and (3) bound-
ing flight (many but not all passerines use this mode of
intermittent flapping with phases of wing folding). Ue
differed significantly between flyers in these three categories
(p , 0.001, adjusted R2 ¼ 0.26, and F2,135 ¼ 25.1), and the
explanatory power of a model incorporating both flight mode
and Q was high (p , 0.001, adjusted R2 ¼ 0.60, and F3,125 ¼
64.5). This suggests that difference in flight mode is one
element affecting the characteristic cruising flight speeds
among phylogenetic groups.
Depending on their ecological life style and foraging, birds
are adapted to different aspects of flight performance, e.g.,
speed, agility, lift generation, escape, take-off, cost of trans-
port, and power [2,10]. These adaptations are likely to have
implications for the flight apparatus (anatomy, physiology,
and muscle operation) and the flight behaviour that may
constrain the cruising flight speed. The variations in power-
Figure 2. Explanation of the Variation in Mean Flight Speeds (Ue; m/s)
versus-speed relationships between different species [22] and
among Bird Species by Different Combinations of Variables and Factors
in muscle efficiency (conversion from metabolic power input
The explanatory power (adjusted R2) of different General Linear Models
with significant independent variables (***, p , 0.001) is illustrated. to mechanical power output) with mass and flight speed
Phylogenetic group and wing loading emerge as key factors to account [23,24] may be related to such differential complex flight
for the variation in flight speed among bird species. General Linear adaptations among birds. Constraints on flight speed may
Models for all different combinations of body mass, wing loading, aspect
ratio, and phylogenetic group were calculated, except combinations also be associated with differences in fluid dynamics and
including both body mass and wing loading (because of the vortex patterns, hereto investigated only for a few species
interdependence between these variables). Complex models (including [25–27]. Variable airspeeds may still be associated with high
combinations of variables) are presented only if the AIC improved from
models based on single independent variables [17]. This applied only to
power efficiency if accompanied with the proper variation in
the model incorporating both phylogenetic group and wing loading. wing stroke frequency and amplitude [28,29].
DAIC indicates the difference in AIC score from the most effective model Species flying at comparatively slow cruising speeds
(with DAIC ¼ 0). Test statistics were as follows (in parentheses) for model frequently use thermal soaring (raptors and storks), are
including mass (F1,136 ¼ 20.0, p , 0.001), aspect ratio (F1,127 ¼ 28.6, p ,
0.001), wing loading (F1,127 ¼ 122.6, p , 0.001), phylogenetic group adapted for hunting and load carrying (raptors), or for take-
(F5,132 ¼ 34.5), and phylogenetic group plus wing loading (F6,122 ¼ 39.6, p off and landing in dense vegetation (herons). Associated with
, 0.001). these flight habits they have a lower ratio of elevator
doi:10.1371/journal.pbio.0050197.g002
(supracoracoideus) to depressor (pectoralis) flight muscle
(particularly low among birds of prey) compared with
explanation for the low scaling exponent of Ue versus Q shorebirds and anatids [2]. We suggest that functional
among bird species as a whole. differences in flight apparatus and musculature among birds
Variability of Flight Speeds of different life and flight styles (differences often associated
with evolutionary origin) have a significant influence on the
Dimensional analyses have demonstrated that scaling
birds’ performance and speed in sustained cruising flight.
relationships between wing loading and total mass differ
Thus, our results strongly indicate that there is a diversity of
significantly between different types of birds [5,10]. The
cruising flight characteristics among different types of birds
expected consequence of this is that wing loading will be a
over and above the general scaling effects of mass and wing
more reliable predictor of flight speed, explaining more of
loading that remains to be investigated and understood,
the variation in flight speeds among bird species than body
aerodynamically [30], kinematically [26,31], physiologically
mass [1,5]. This expectation was fully confirmed in the
[22], as well as ecologically [2,10].
present study, with Q accounting for almost half of the
variation in Ue between species, while m explained only 12%
of this variation (Figure 2). However, our findings that Q still Materials and Methods
left a large part of the variation in flight speed unexplained Tracking radar measurements. Our main dataset, based on
and that phylogenetic group accounted for a significant tracking radar measurements in Sweden and the Arctic 1979–1999,
fraction of this remaining variation were unexpected from consists of 1,399 tracks of 102 identified species, with a mean track
time of 369 s (range 20–2,220 s). Altitudes ranged from sea level to
earlier analyses based on theoretically calculated flight speeds 3,600 m. Number of tracks for each species ranged between one and
[5,10]. 240, and mean Ue (with SD), vertical speed as well as information
What are the causes for the discrepancies in flight speed about number of tracks, track time, and biometry data are given for
each species in Protocol S1.
between phylogenetic groups? Differences in flight mode and An extensive additional dataset of equivalent airspeeds of
the use of bounding flight by many passerines have been identified birds, obtained by similar tracking radar techniques, has
suggested as explanations for important group-specific been published from the work of Bruno Bruderer and his research
deviations from aerodynamic predictions of optimal bird group in Switzerland, Germany, Israel, and Spain [15]. Flight speed
data from tracks of birds in natural migratory flight (excluding
flight speeds [15]. We provisionally assigned, based on our released birds and soaring flight) were incorporated into our analysis.
own field experience, the different bird species to three main This additional dataset comprised 64 species, and with 28 species

PLoS Biology | www.plosbiology.org 1660 August 2007 | Volume 5 | Issue 8 | e197

 Bird Flight Speeds

shared between the two sets of data, the combined data added up to a phylogenetic group) were presented in Figure 2 only if AIC improved
total of 138 species (Protocol S1). Mean Ue for the shared species were from that of models with single independent variables [19].
not significantly different between the two sets (paired sample t-test, t
¼ 1.28, and p ¼ 0.21), and we used weighted (according to the number
of tracks) overall mean Ue for these species in our analyses. Supporting Information
The bulk of flight speed data were measured 1979–1999 by tracking
radar studies at five sites in southern Sweden and on two expeditions Protocol S1. Supplementary List of Flight Speeds and Biometry of
by icebreaker to the Arctic (for detailed methods see [19,32]). Targets Bird Species
were identified to species and flock sizes through telescopes Found at doi:10.1371/journal.pbio.0050197.sd001 (173 KB PDF).
simultaneously with radar registrations providing computer readings
of range, elevation, and bearing to the target usually every 10 s with Protocol S2. Supplementary Information on Phylogenetic Tree,
the radar in automatic tracking mode. All flight speeds have been Taxon Sampling, and Analysis of Independent Phylogenetic Con-
corrected for the influence of wind by subtraction of the wind vector trasts
at the altitude where the birds were flying from the ground speed Found at doi:10.1371/journal.pbio.0050197.sd002 (29 KB PDF).
vector of the birds. Winds were measured by releasing and tracking
hydrogen/helium-filled balloons carrying a radar reflector. Mean
airspeed, altitude, and vertical flight speed were calculated for each Acknowledgments
track, excluding segments with a convoluted flight path. Altitudes
were corrected in relation to sea level by adding the altitude of the We are very grateful to Inga Rudebeck who participated in all radar
radar antenna (10–185 m above sea level at the different sites), and fieldwork in Sweden and calculated and compiled the radar tracking
true airspeeds were reduced to equivalent airspeeds (Ue) referring to results and to Bertil Larsson who supervised the radar operation,
sea level air density, according to the standard atmosphere change in equipment, and software for registration. We also thank several
air density with altitude [14,15]. additional participants in the field work: M. Green, G. A. Gudmunds-
Scaling calculations and statistical analyses. Reduced major axis son, A. Hedenström, and A. Ulfstrand. Radar services and recon-
regressions [16] for the scaling relationships between Ue and m and Q, structions were made by Aerotech Telub. Icebreaker expeditions to
respectively, were performed in Matlab, with calculations of the Arctic were organized by the Swedish Polar Research Secretariat.
confidence intervals by bootstrapping [33]. Calculations of reduced We thank B. Bruderer and A. Hedenström for comments on the
major axis regressions based on phylogenetic independent contrasts manuscript and L. Larsson and M. Irestedt for discussions about the
are further described in Protocol S2. We checked for possible bias taxonomic classification of birds. We are also grateful to Theunis
arising as a consequence of including species with only one or a few Piersma and anonymous referees for valuable comments and
tracks, by restricting the calculations to species with at least five or suggestions.
ten tracks. The results remained the same, as exemplified for the Author contributions. TA and his laboratory organized and carried
sample of 56 species with 10 tracks in Table 1. For 39 of the species out the radar field work. MR extracted and prepared the final dataset
with 10 tracks, we could account for the within-species variation of and performed the scaling analyses. JB organized the tracking
Ue in relation to vertical flight speed, head- and side-wind information into a database, including information about biometry.
components, and flock size by multivariate regression (statistically PGPE constructed the phylogenetic tree and classified the species
significant influences were found in 26 of these 39 species; into main phylogenetic groups, and OH performed the scaling
unpublished data). Restricting the analysis to intercept values of Ue analyses based on phylogenetic contrast data. TA, MR, JB, PGPE, and
for these 39 species (corrected to zero vertical speed, zero wind, and a OH participated in the evaluation and discussion of results and
flock size of one from the multiple regression equations of significant writing of manuscript.
variables for each species) still gave the same scaling result (Table 1). Funding. This work was funded by grants from the Swedish Natural
General Linear Models (Figure 2) [34] were calculated with Ue as Science Research Council and the Swedish Research Council to TA.
dependent variable. Logarithmic values were used for Ue, m, and Q. Reconstruction of the radar for bird tracking purposes was financed
Phylogenetic group and flight mode (limited analysis of this provi- by grants from Knut and Alice Wallenbergs Foundation and the
sionally estimated variable) were treated as fixed factors. Complex Swedish Council for Planning and Coordination of Research.
models (different combinations or interactions of mass, aspect ratio, Competing interests. The authors have declared that no competing
and phylogenetic group or of wing loading, aspect ratio, and interests exist.

References 15. Bruderer B, Boldt A (2001) Flight characteristics of birds: I. Radar

1. Lighthill J (1977) Introduction to the scaling of animal locomotion. In: measurements of speeds. Ibis 143: 178–204.
Pedley TJ, editor. Scale effects of animal locomotion. New York: Academic 16. Harvey PH, Pagel MD (1991) The comparative method in evolutionary
Press. pp, 365–404. biology. Oxford: Oxford University Press. 239 p.
2. Rayner JMV (1988) Form and function in avian flight. In: Johnston RF, 17. Johnson JB, Omland KS (2004) Model selection in ecology and evolution.
editor. Volume 5, Current ornithology. New York and London: Plenum Trends Ecol Evol 19: 101–108.
Press. pp. 1–66. 18. Pennycuick CJ (1975) Mechanics of flight. In: Farner DS, King JR, editors.
3. Spedding GR (1992) The aerodynamics of flight. In: Alexander R, editor. Volume 5, Avian biology. New York: Academic Press. pp. 1–75.
Volume 11, Advances in comparative and environmental physiology. Berlin 19. Hedenström A, Alerstam T (1992) Climbing performance of migrating
and Heidelberg: Springer-Verlag. pp. 52–111. birds as a basis for estimating limits for fuel carrying capacity and muscle
4. Pennycuick CJ (1969) The mechanics of bird migration. Ibis 111: 525–556. work. J Exp Biol 164: 19–38.
5. Greenewalt CH (1975) The flight of birds. Volume 65, Trans Am Phil Soc 20. Vogel S (1994) Life in moving fluids. The physical biology of flow. 2nd
New Series-part 4. 67 p. edition. Princeton: Princeton University Press. 467 p.
6. Tennekes H (1997) The simple science of flight - from insects to jumbo jets. 21. Pennycuick CJ (2001) Speeds and wingbeat frequences of migrating birds
Cambridge (Massachusetts): MIT Press. 138 p. compared with calculated benchmarks. J Exp Biol 204: 3283–3294.
7. Tucker VA (1973) Bird metabolism during flight: Evaluation of a theory. J 22. Tobalske BW, Hedrick TL, Dial KP, Biewener AA (2003) Comparative
Exp Biol 58: 689–709. power curves in bird flight. Nature 421: 363–366.
8. Rayner JMV (1979) A new approach to animal flight mechanics. J Exp Biol 23. Kvist A, Lindström Å, Green M, Piersma T, Visser GH (2001) Carrying large
80: 17–54. fuel loads during sustained flight is cheaper than expected. Nature 413:
9. Rayner JMV (1995) Flight mechanics and constraints on flight performance. 730–732.
Isr J Zool 41: 321–342. 24. Ward S, Möller U, Rayner JMV, Jackson DM, Bilo D, et al. (2001) Metabolic
10. Norberg UM (1990) Vertebrate flight. Berlin: Springer-Verlag. 291 p. power, mechanical power and efficiency during wind tunnel flight by the
11. Pennycuick CJ (1982) The flight of petrels and albatrosses (Procellariiformes), European starling Sturnus vulgaris. J Exp Biol 204: 3311–3322.
observed in South Georgia and its vicinity. Phil Trans R Soc London B 300: 25. Spedding GR, Rosén M, Hedenström A (2003) A family of vortex wakes
75–106. generated by a thrush nightingale in free flight in a wind tunnel over its
12. Ericson PGP, Anderson CL, Britton T, Elzanowski A, Johansson US, et al. entire natural range of flight speeds. J Exp Biol 206: 2313–2344.
(2006) Diversification of Neoaves: Integration of molecular sequence data 26. Rosén M, Spedding GR, Hedenström A (2004) The relationship between
and fossils. Biol Let 2: 543–547. wingbeat kinematics and vortex wake structure in a thrush nightingale. J
13. Hedenström A, Alerstam T (1995) Optimal flight speed of birds. Phil Trans Exp Biol 207: 4255–4268.
R Soc Lond B 348: 471–487. 27. Hedenström A, Rosén M, Spedding GR (2006) Vortex wakes generated by
14. Pennycuick CJ (1989) Bird flight performance: A practical calculation robins Erithacus rubecula during free flight in a wind tunnel. J R Soc
manual. Oxford: Oxford University Press. 153 p. Interface 3: 263–276.

PLoS Biology | www.plosbiology.org 1661 August 2007 | Volume 5 | Issue 8 | e197

 Bird Flight Speeds

28. Taylor GK, Nudds RL, Thomas ALR (2003) Flying and swimming animals 31. Tobalske BW, Hedrick TL, Biewener AA (2003) Wing kinematics of avian
cruise at a Strouhal number tuned for high power efficiency. Nature 425: flight across speeds. J Avian Biol 34: 177–184.
707–711. 32. Alerstam T, Gudmundsson GA (1999) Migration patterns of tundra birds:
29. Nudds RL, Taylor GK, Thomas ALR (2004) Tuning of Strouhal number for Tracking radar observations along the Northeast Passage. Arctic 52: 346–
high propulsive efficiency accurately predicts how wingbeat frequency and 371.
stroke amplitude relate and scale with size and flight speed in birds. Proc R 33. Bohonak AJ (2004) RMA, software for reduced major axis regression
Soc Lond B 271: 2071–2076. version 1.17 [computer program]. Available: http://www.bio.sdsu.edu/pub/
30. Spedding GR, Hedenström A, Rosén M (2003) Quantitative studies of the andy/rma.html. Accessed 1 July 2006.
wakes of freely flying birds in a low-turbulence wind tunnel. Exp Fluids 34: 34. SPSS (2003) Version 12.0.1 for Windows [computer program]. Available:
291–303. http://www.spss.com. Accessed 1 November 2003.

PLoS Biology | www.plosbiology.org 1662 August 2007 | Volume 5 | Issue 8 | e197

View publication stats