The Aerodynamics of Hummingbird Flight

Abstract

(Abstract) Hummingbirds fly with their wings almost fully extended during their entire wingbeat. This pattern, associated with having proportionally short humeral bones, long distal wing elements, and assumed to be an adaptation for extended hovering flight, has lead to predictions that the aerodynamic mechanisms exploited by hummingbirds during hovering should be similar to those observed in insects. To test these predictions, we flew rufous hummingbirds (Selasphorus rufus, 3.3 g, n = 6) in a variable-speed wind tunnel (0-12 ms-1) and measured wake structure and dynamics using digital particle image velocimetry (DPIV). Unlike hovering insects, hummingbirds produced 75% of their weight support during downstroke and only 25% during upstroke, an asymmetry due to the inversion of their cambered wings during upstroke. Further, we have found no evidence of sustained, attached leading edge vorticity (LEV) during up or downstroke, as has been seen in similarly-sized insects - although a transient LEV is produced during the rapid change in angle of attack at the end of the downstroke. Finally, although an extended-wing upstroke during forward flight has long been thought to produce lift and negative thrust, we found circulation during downstroke alone to be sufficient to support body weight, and that some positive thrust was produced during upstroke, as evidenced by a vortex pair shed into the wake of all upstrokes at speeds of 4 - 12 m s-1.

Full text

American Institute of Aeronautics and Astronautics
1
The Aerodynamics of Hummingbird Flight
Douglas R. Warrick
*
and Bret W. Tobalske.
†
Oregon State University, Corvallis Oregon 97331 and University of Portland, Portland OR 97203
Donald R. Powers
‡
George Fox University, Newburg, OR 97132
and
Michael H. Dickinson
§
California Institute of Technology, Pasadena, CA 91125
[Abstract] Hummingbirds fly with their wings almost fully extended during their entire
wingbeat. This pattern, associated with having proportionally short humeral bones, long
distal wing elements, and assumed to be an adaptation for extended hovering flight, has lead
to predictions that the aerodynamic mechanisms exploited by hummingbirds during
hovering should be similar to those observed in insects. To test these predictions, we flew
rufous hummingbirds (Selasphorus rufus, 3.3 g, n = 6) in a variable–speed wind tunnel (0-12
ms
-1
) and measured wake structure and dynamics using digital particle image velocimetry
(DPIV). Unlike hovering insects, hummingbirds produced 75% of their weight support
during downstroke and only 25% during upstroke, an asymmetry due to the inversion of
their cambered wings during upstroke. Further, we have found no evidence of sustained,
attached leading edge vorticity (LEV) during up or downstroke, as has been seen in
similarly-sized insects - although a transient LEV is produced during the rapid change in
angle of attack at the end of the downstroke. Finally, although an extended-wing upstroke
during forward flight has long been thought to produce lift and negative thrust, we found
circulation during downstroke alone to be sufficient to support body weight, and that some
positive thrust was produced during upstroke, as evidenced by a vortex pair shed into the
wake of all upstrokes at speeds of 4 – 12 m s
-1
.
I. Introduction
ITH a few exceptional intersections, the evolution of human-engineered flight and the study of the evolution
of animal flight have been essentially parallel. Given the results of the earliest such meetings (e.g.,
DaVinci’s ornithopter), this has probably been for the best; the disparity in scale between these lineages and
its effects on structural and fluid mechanics has necessarily cloistered these two fields and prevented further fruitless
and dangerous intercourse. However, interest in the development of micro-air-vehicles (MAVs) has thrown a
debutante ball, and it would seem that the convergence, the meeting and mixing of these lines – now working at
similar scales and Reynolds numbers (Re) – could produce useful offspring. The key to the viability of such
products will be determining which of those characteristics described for biological fliers are results of natural
selection, rather than results of ancestry. That is, the utility of our understanding of biological flight to the
engineering community rests upon our ability to determine adaptation – a question fundamental to biologists.
Certainly, some of the loveliest of biological models to walk onto the dance floor are the hummingbirds. Possessing
the right range of sizes (from 2-20 grams), unmatched aerial performance for animals of those sizes, along with
important research intangibles (i.e., tractability and warmth), hummingbirds seem likely sources of useful design
*
Assistant Professor, Department of Zoology, 3029 Cordley Hall.
†
Associate Professor, Department of Biology, 5000 N Willamette Blvd.
‡
Professor and Chair, Department of Biology, 414 N. Meridian Street.
§
Esther M. and Abe M. Zarem Professor of Bioengineering, Division of Biology, 1200 E California Blvd.
W

American Institute of Aeronautics and Astronautics
2
seeds – places for engineers to start. The purpose of this paper is to describe what is currently known of the
biomechanics and aerodynamics of hummingbirds, placed in an evolutionary context that will allow those who
might use these animals as models to assess those mechanisms as worthy of emulation.
II. Hovering Flight: Lessons from Convergence?
Thirty-three years after the first rigorous treatment of
hummingbird flight by Stolpe and Zimmer in 1939
1
, describing
the kinematics of hovering, Weis-Fogh (1972)
2
developed an
aerodynamic model of hummingbird flight, noting its
similarities to insect flight. Although the fruit flies operated at
Re considerably lower than those of the hummingbirds, their
kinematic similarities led to the assumption
2,3,4,
that the
aerodynamic mechanisms were also similar – most notably in
that the two halves of the wingbeat cycle were roughly similar
in aerodynamic force production. More recently, flow
visualization and dynamically-scaled robotic simulations of
insect flight
5,6
demonstrated that the half-strokes of insects
similar in size to hummingbirds, were indeed aerodynamically
active and equal, and that lift was generated through leading-
edge vorticity (LEV) attached to the dorsal surface of the
translating wing. Hummingbird wings, modeled as flat plates
and flown in a dynamically-scaled robot, produced similar
LEVs and symmetrical force when flown at kinematic angles of
attack observed in hovering hummingbirds (Fig.1). The
inferences it made to hummingbird flight were clear and
compelling; the demonstration of convergence, from two such
long-diverged evolutionary lines, on the same locomotor
mechanism would make a profound statement to MAV
engineers: there’s only one way to hover a small, flapping
vehicle.
We tested the assumptions of half-stroke symmetry using 2-D
digital particle image velocimetry (DPIV), sampling the wake
of hummingbirds in both the frontal and parasagittal plane (see
Appendix) allowing us to capture tip and starting/ending
vorticity respectively. We found that hummingbirds exhibit
marked asymmetry in lift force production, with 75% of the
body weight lift support being generated during the downstroke
– in essence, about half-way from the typical bird condition
(100% of lift during the downstroke) and the typical insect
7
(Fig.2). Given the respective wing speeds and areas of the two
half-strokes, we concluded that this asymmetry was at least in
part due to the positive camber of the hummingbird wing,
which does not completely reverse during upstroke. However,
the far-field wake, remarkably similar to that produced by
hawkmoths
6
(Fig. 3) did suggest that LEVs were present during
at least the downstroke, and that hummingbirds might be true
hovering chimeras. While it seemed certain that leading-edge
vorticity was formed at some point during the downstroke, the
question of the intent and purpose of such flow remained.
Preliminary examination of the near-field flow around
hummingbird wings during both half-strokes reveals that no
attached, stable leading-edge vorticity is developed during the
Figure 2. Frontal plane vorticity (Rot Z [dx/dy]) field
of a hovering hummingbird. Upper vortex pair (a) are
tip vortices from upstroke; lower pair were produced
by the previous downstroke.
Figure 3. Parasagittal plane vorticity. (Rot Z
[dx/dy]) field of a hovering hummingbird. Upper
vortex (a) is the starting vortex of upstroke; (b) putative
LEV, produced by the rapid pitch up of the wing; (c)
ending vortex of downstroke.
Figure 1. DPIV vector field. Flow around a
dynamically-scaled robotic simulation of hummingbird
wings. Wings were modeled as flat plates; angle of
incidence = 15
o
. Note clockwise attached LEV above
wing.
b. 8282300 s
-1
a. 2827130 s
-1
8718780 s
-1
1005040 s
-1
b. 8282300 s
-1
a. 2827130 s
-1
8718780 s
-1
1005040 s
-1
a.
b.
c.
a.
b.
c.

American Institute of Aeronautics and Astronautics
3
majority of the wing translation (Fig. 4); that is, the flow is essentially laminar, and typical lifting line aerodynamics
probably explain the majority of lift production. The vorticity seen in the far-field wake would seem to be a result
of a transient LEV produced by the rapidly-pitching airfoil at the end of the downstroke – an effect typical of a
dynamically-stalling airfoil. However, the flow at the leading edge of the wing is not entirely typical; in particular,
the stagnation point of the airflow is deep – several millimeters back from the leading edge on the ventral side of the
wing. The movement of air from this ventral point, around the leading edge to the dorsal surface of the wing
ostensibly creates a vortex with a center at the anatomical leading edge of the wing. In outward appearance, this
flow is similar to that around the leading edge of fixed (i.e, gliding) swift wings
8
, but due to the differences in wing
presentation (strongly swept in the swift versus straight in the hummingbird) the equivalence of these structures
cannot be determined.
The observed difference in flow between the robotic model and
real hummingbird wing (Figs. 1, 4) may have several origins;
the two most immediately recognizable are both a result of the
effects of camber. As previously suggested, the greater lift
coefficients generated by cambered wing of the real bird
generated greater downwash during downstroke, forming the
basis of half-stroke disparity. This greater downwash also
lowers the effective angle of attack; the 25
o
-30
o
angle of
incidence in early downstroke is reduced to 9
o
-15
o
when a near-
field downwash of 3 ms
-1
is incorporated into the calculation of
angle of incidence. The flow around a flat plate at this angle
may be more similar to the observed flow around hummingbird
wings.
The aerodynamic properties of feathers cannot be discounted as
a source of the unique flow around hummingbird wings, and
studies of the effects of the microstructure of feathers on
boundary-layer interactions – either through surface ‘tripping’
or airfoil transmissivity, are probably warranted.
III. Forward flight
While the advantages of leaving the wing
extended during the upstroke during
hovering are clear, there are good
theoretical reasons
4,9
, and some
experimental evidence
10,11
to suggest it
may be a liability at low forward flight
speeds and during acceleration. During
upstroke, the supinated wing should
produce a lift force with a rearward vector
component, producing negative thrust,
adding to the profile drag already produced
by the wing. Thus, most small birds flex
their wings during the upstroke, completely
ceasing lift production and avoiding these
aerodynamic penalties
12
– which, for small
birds, with relatively low inertia, may be
severe. Perhaps owning to the anatomical
commitment to hovering flight,
hummingbirds leave their wings extended
during upstroke at all flight speeds. The
far-field wake of hummingbirds indicates
that lift production indeed continues
-20 0 0
2000
Vort icit y (s
-1
)
5cm
Downstroke
Down
a
b
c
d
Figure 5. Wake of a hummingbird. Forward flight, (6 ms
-1
). Note the
downstroke starting (a) and ending vortices (b), and the vortex pair,
hypothetically a result of a secondary ‘post-ending’ vortex at mid-
upstroke (c )and a ‘pre-starting’ vortex (d) produced in quick
succession. The interaction of this vortex pair suggests thrust
production.
Figure 4. Flow around a hummingbird wing.
DPIV vector field of flow around a hummingbird
mid-wing at mid-downstroke. Red line indicates the
position of the chord x-section, moving left to right.
Note the absence of dynamic stall vorticity, present
in the robotic simulation (Fig. 1), above the wing.

American Institute of Aeronautics and Astronautics
4
throughout the upstroke; however, there is little to suggest that negative thrust is produced during this type of
upstroke. Indeed, the shedding of a vortex pair at mid-to-late upstroke, coinciding with the point in the upstroke
where the speed of the upward translation of the wing has reduced angle of incidence to zero (Fig. 5), suggests that
some thrust may be generated during this portion of the wingbeat cycle. The strength and utility – and ubiquity
among other birds with aerodynamically active upstrokes - of this mechanism to hummingbirds have yet to be
determined, but it is probably insufficient to overcome the aerodynamic costs of active upstrokes in birds (or MAVs)
with larger wings. Evidence from the anatomy and performance of swifts
10
suggests that unusually large pectoral
muscles (and, hence, particularly strong downstrokes for thrust production) are required to overcome the costs
incurred by stiff wings and active upstrokes.
IV. Conclusion
DPIV analysis of live birds and robotic simulations suggests that for thin wings at low Re, cambered airfoils
generate greater lift coefficients than flat plates (or other such symmetrical sections), and it seems likely that the
asymmetry in the half-strokes of hovering hummingbirds is clearly an artifact of its avian ancestry – ancestors for
which the downstroke (or just a wing extended in glide) was the only lift generating portion of the wingbeat cycle.
Given that symmetry in lift production offers some advantages (e.g., more continuous availability of lift force for
needs of maneuvering; smaller vertical oscillations in body movement between the half-strokes), there would be
little use in incorporating this particular aspect of hummingbird flight into a MAV. However, emulating the rigid
and kinematically simple wing of hummingbirds may be extremely useful, given its performance over a range of
speeds.
Appendix
A. Circulation and Weight Support
To compute vorticity (ω, s
-1
), we post-processed
vector fields using a median filter, and then computed
rot z[dy/dx]. We measured circulation (Γ, m
2
s
-1
) in the
trailing tip vortices by integrating ω with respect to area
(m
2
). We limited our analysis to views where vortex
cores were normal to the sampling plane (parasagittal:
centered at midwing; frontal: centered at wing root). We
tested whether observed Γ was sufficient to support
body weight by comparing Γ with circulation required
(Γ
ο
) = WT / ρS where W is body weight (N), T is time
per wingbeat (s), and S is the projected horizontal area
swept by the two wings (m
2
)
13
.
B. Kinematics
Separate flight trials (n = 4 birds) were
recorded using two synchronized high speed digital
video cameras operating at 500 Hz sampling and shutter
speed of 1/1000 s. We merged two-dimensional
coordinates from each camera into a single 3-d
coordinate space using the direct linear transformation
(DLT) coefficients derived from a 16-point calibration
frame
14
. Using these data, we calculated angular
velocity of the wing (rad s
-1
) and angle of attack of the
mid-wing (degrees) relative to incurrent air flow.
Incurrent air velocity was the sum of translational
Figure 6. Position of Hummingbirds hovering. DPIV
laser light sheet illustrating the frontal (a,b) and
parasagittal (c,d) sampling planes and revealed wake
structures.

American Institute of Aeronautics and Astronautics
5
velocity of the wing and average 3-d air velocity computed using DPIV data from frontal and sagittal planes, which
is dominated by a mean downward velocity of 1.1 ms
-1
.
Acknowledgments
We thank Dr. Andrew Johnson, AHPCRC, Dr. Steve Anderson, LaVision Inc., Dr. Ty Hedrick, UNC Chapel
Hill, Beth Klopfenstein, George Fox University, Gabe Hyder and John Ranola, University of Portland, Will
Dickson, California Institute of Technology, and Doug Altshuler, UC Riverside. Supported by NSF IBN 0327380
and IOB 0615648.
References
1
Stolpe, V. M. & Zimmer, K. “Der schwirrflug des kolibri im zeitlupenfilm,” J. feur Ornith. Vol. 87, 1939, pp. 136-155
2
Weis-Fogh, T. “Energetics of hovering flight in hummingbirds and in Drosophila,” J. Exp. Biol. Vol. 56, 1972, pp. 79–104
3
Greenwalt, C. H. “The wings of insects and birds as mechanical oscillators,” Proc. Amer. Phil. Soc. Vol. 104, 1960, pp.
605-611
4
Norberg, U. M. Vertebrate Flight: Mechanics, Physiology, Morphology, Ecology, And Evolution. Springer-Verlag, Berlin
1990
5
Wilmott, A. P. & Ellington, C. P. “The mechanics of flight in the hawkmoth Manduca sexta II. Aerodynamic consequences of
kinematic and morphological variation,” J. Exp. Biol. Vol. 200, 1997,
pp. 2723-2745
6
van den Berg, C., & Ellington, C. P. “The vortex wake of a ‘hovering’ model hawkmoth,” Phil. Trans. R. Soc. Lond. B Vol.
352, 1997,
pp. 329–340
6
Ellington, C. P. “The aerodynamics of hovering insect flight. V. A vortex theory,” Phil. Trans. R. Soc. Lond. B Vol. 305,
1984,
pp. 79-113
7
Warrick, D. R., , Tobalske, B. W., and Powers, D. R. “Aerodynamics of the hovering hummingbird,” Nature, Vol. 435, 2005,
pp.1094-1097.
8
Videler, J. J., Stamhuis, E. J., and Povel, G. D. E. “Leading-edge vortex lifts swifts,” Science, Vol. 306, 2004, pp. 1960-1962.
9
Spedding, G. R., “The wake of a kestrel (Falco tinnunculus) in flapping flight,” J. Exp. Biol., Vol. 127, 1987, pp. 69-78.
10
Warrick, D. R. “The turning and linear maneuvering performance of birds: the cost of efficiency for coursing insectivores.
Can. J. Zool. Vol. 76, 1998, pp. 1063-1079.
11
Hedrick, T. L., Usherwood, J. R., and Biewener, A. A. “Wing inertia and whole-body acceleration: an analysis of
instantaneous dynamic force production in cockatiels (Nymphicus hollandicus) flying across a range of speeds.” J. Exp. Biol.
Vol. 207, 2002, pp. 1689-1702.
12
Tobalske, B. W., Hedrick, T. L. & Biewener, A. A. “Wing kinematics of avian flight across speeds.,” J. Avian Biol. Vol. 34,
2003,
pp. 177-184
13
Spedding, G. R., Hendenstrom, A. & Rosen, M. “Quantitative studies of the wakes of freely flying birds in a low-turbulence
wind tunnel,” Exp. Fluids Vol. 34, 2003,
pp. 291-303
14
Hedrick, T. L., Tobalske, B. W. & Biewener, A. A. “Estimates of gait change based on a three-dimensional analysis of flight in
cockatiels (Nymphicus hollandicus) and ringed turtle doves (Stretopelia risoria).,” J. Exp. Biol. Vol. 205, 2002,
pp. 1389-1409