Kinematics and power requirements of ascending and descending flight in the pigeon (Columba livia)

While in flight, a bird often needs to adjust its altitude. Changing altitude may be necessary to land, forage, pursue prey, or to maneuver through the environment. As with terrestrial locomotion, an increase in elevation during flight results in a change in the potential energy of the animal's center of mass. But unlike terrestrial animals, a bird in flight cannot push against a solid substrate to change its altitude. Instead, a bird must depend on interactions between its wings, tail and body and the air to effect this change.

Whereas the mechanics and energetics of incline locomotion have been fairly well studied in terrestrial animals (e.g. Gillis and Biewener, 2002; Dutto et al., 2004; Higham and Jayne, 2004),including humans (Iversen and McMahon,1992; Gottschall and Kram,2005; Roberts and Belliveau,2005), and birds performing terrestrial locomotion(Bundle and Dial, 2003; Daley and Biewener, 2003; Gabaldon et al., 2004; Rubenson et al., 2006), there has been relatively little focus on the ascending and descending flight of birds. Past work has examined in vivo muscle function of birds during vertical or angled take-off and landing(Dial, 1992; Dial and Biewener, 1993; Askew et al., 2001). However,the kinematics and aerodynamic mechanisms underlying steady ascending and descending avian flight have remained largely unexplored.

In contrast to weight support during terrestrial locomotion, a bird, even when hovering and remaining stationary, must continually force air downward with its wings to support its weight (e.g. Norberg, 1990). Staying in place during flapping flight thus requires mechanical power. Freely hovering hummingbirds, for example, show a whole-body power output ranging from 22 to 38 W kg^–1 (Altshuler et al., 2004). Steady, level, forward flight likewise requires power output, though there is no net change in kinetic or potential energy over the course of a wingbeat. Many studies have explored the relationship between forward flight speed and power output, with the goal of assessing whether birds display a U-shaped power curve and how differences in wing shape and flight style might affect this relationship (e.g. Torre-Bueno and Larochelle,1978; Rothe et al.,1987; Rayner,1999; Tobalske et al.,2003; Bundle et al.,2007).

Another straightforward way to explore differences in power output is to require a bird to change its potential energy by having it fly along an ascending or descending path. The main goal of the present study was to begin to explore how flight power requirements change as a function of flight path angle, by examining the free flight of pigeons trained to fly over varying angles of ascent and descent. In particular, we sought to determine whether the power required for incline or decline flight could be measured by simply summing the power necessary for level flight at the same speed and the rate of center of mass (CoM) potential energy change. Given our hypothesis that this would be the case, we expected to find that, compared to level flight,ascending flight would require more power, descending flight would require less power, and the difference in each case would be the rate of CoM potential energy change.

We also sought to explore how the direction and magnitude of average aerodynamic force produced by the wings during downstroke changes with flight angle. Relative to level flight, we expected that aerodynamic force would be greater or rotated upward for climbing flight, as either alteration would result in an increased upward component of aerodynamic force.

By studying the kinematics of pigeons during ascending and descending flight, we sought to determine patterns of wing motion that might underlie differences in power and force between level and angled flight. We expected angle of attack to be determined primarily by the orientation of the incident velocity, which is largely influenced by the angle of flight. We thus expected to observe lower angles of attack for ascending flights and greater angles of attack for descending flights. Given that ascending flight would require a greater power output and descending flight a lower power output than level flight at a given speed, we also expected to observe greater wingbeat frequency for ascending flights, and lower wingbeat frequency for descending flights. We did not expect flight speed or stroke plane angle to vary among flight angles, as the size of our experimental flight arena did not allow for fast flight. Because of the relatively slow flight speed used by the pigeons in our experiments, the wingstroke would primarily need to direct airflow downward to support the bird's weight, suggesting that the stroke plane angle would be nearly horizontal for all conditions.

Four rock pigeons Columba livia Linnaeus (hereafter `pigeons'; Table 1) from a small flock housed at the Concord Field Station were used in the experiments. These pigeons were bred on-site from carrier pigeon stock. Birds were provided with food and water ad libitum. Body mass and wing data for each bird are shown in Table 1. Two wooden perches (one 22×62×2 cm and another slightly larger,30×50×2 cm, covered with cardboard, thin plastic and paper towels)were set up in an arena (8 m long×2 m wide×4 m high) enclosed with lightweight fruit netting to prevent the pigeons from escaping. Pigeons were trained regularly (∼30 min per day) for several weeks to fly between the perches (Fig. 1A). The heights of both perches were adjustable to permit control of the angle of flight. The slightly larger perch was elevated and locked into position on a 5 cm diameter PVC support pipe anchored at the ceiling and floor. The perch heights were adjusted incrementally, and the pigeons were continually encouraged to fly between them until the desired flight angle was reached(Table 2). The pigeons were conditioned to fly at each of the flight angles to be studied (60°,30°, 0°, –30° and –60°). The actual angles of flight observed during the experiment differed slightly from the angle of the setup (see Results). For ease of discussion, however, these conditions will still be referred to as 60°, 30°, etc. The terms `steep' and `shallow'will be used to refer to 60°/–60° and 30°/–30°flights, respectively.

Before obtaining high-speed video recordings for 3D kinematic analysis (see below), pigeons were marked at several anatomical landmarks(Fig. 1B). These included the left and right shoulders, wrists, longest primary feather on each wing(usually the ninth), and the rump. In the global coordinate system, the x-axis was defined as being horizontal, in line with the perch supports; the y-axis was medio-lateral; and the z-axis was vertical.

Three high-speed, digital video cameras (two RedLake PCI 500, RedLake Inc.,San Diego, CA, USA; and one Photron FastCam-X 1280 PCI, Photron USA Inc., San Diego, CA, USA) were positioned around the flight arena. One RedLake camera was mounted on the ceiling to record dorsal views of the pigeons. The other RedLake camera was positioned near the taller perch, and the Photron camera was positioned to the side of the arena(Fig. 1A). Recordings were made at 250 frames s^–1, with shutter speeds of 1/500 to 1/1000 s. The flight volume from which 3D kinematics were taken (XYZ:1×0.9×1 m for level flight; 1×0.9×2 m for angled flight) was calibrated using the direct linear transform method(Hatze, 1988). The calibrated volume included approximately 13% of the level flight path, 19% of the±30° flight paths and 56% of the ±60° flight paths.

At least ten flights were recorded for each bird at each flight angle, and in general, four of these were used for analysis. Flights were selected for analysis based on the straightness of the bird's path, as viewed in the dorsal camera. The dorsal camera view was the most restricted of the three camera views, and only recordings that contained at least one full wingbeat in the dorsal camera view were used for analysis. A wingbeat was defined as the movement from one upstroke–downstroke transition to the next. At–30°, no recordings for Bird 1 met this criterion, so the results presented for –30° flight reflect only data from Birds 2, 3 and 4. Birds typically flew between the two perches using ∼9 wingbeats for 0°flights, ∼13 wingbeats for 30° ascents, ∼11 wingbeats for–30° descents, ∼11 wingbeats for 60° ascents, and ∼8 wingbeats for –60° descents. For each flight, the wingbeat analyzed was the one that occurred closest to the middle of the dorsal camera view. This ranged from the third to the fifth-to-last wingbeat.

Kinematic marks on the birds were digitized using the custom MatLab(Version 6.5 Release 13; 2002, The MathWorks, Inc., Natick, MA, USA) program DLTdataviewer, written by T. Hedrick(Hedrick et al., 2004). Because the tips of the feathers on the trailing edge of the wing change position dramatically during a wingstroke, kinematic markers could not easily be used to give the angle of the wing at a particular distance along the wing. Consequently, the point on the trailing edge directly behind the wrist in the parasagittal plane was digitized. The position of the center of mass (CoM) of the bird was estimated by averaging the x- and z-coordinates of the rump and shoulders, and using the y-coordinate of the rump(Fig. 1B). Although the CoM of the bird's body will change during a wingbeat cycle, largely due to motion of the wings, this effect is minimal in comparison with changes in CoM position over the course of the complete wingbeat cycle. However, these time-varying inertial effects do influence accelerations during the wingbeat(Hedrick et al., 2004). Because we were observing non-maneuvering flights, we assumed that inertial effects would not create a systematic difference in mid-downstroke calculations across flight angles.

For each digitized wingbeat, the kinematic data were filtered at four times the wingbeat frequency (∼30 Hz) with a low-pass fourth-order Butterworth filter implemented in a custom MatLab script. The filtered wingbeats for each bird were normalized to the same duration and then averaged within each condition. This procedure produced the standard wingbeats used for analysis. Instantaneous velocities and accelerations were calculated by numerical differentiation of the filtered positional data. Flight velocity Vwas calculated as the resultant of the velocities measured in each of the global x, y and z directions.

Half-stroke transitions were determined by selecting minima for the y-coordinate of the wingtip, i.e. when the wingtip was most medial,relative to the shoulder joint of the wing. Mid-half-stroke points were determined by selecting maxima in the y-coordinate of the wingtip,i.e. when the wingtip was farthest from the shoulder joint. We defined`mid-downstroke' as the middle 16 ms (five video frames) of the downstroke. Whole wingbeats were defined as the movement from one upstroke–downstroke transition to the next.

The stroke plane was determined by performing a linear regression of the x- and z-coordinates of the kinematic data for the wingtip,relative to the shoulder, with the slope providing a measure of the stroke plane angle (Fig. 2). The maximum angle of excursion of the wing, Φ, was determined by finding the maximum angle between shoulder and paired wingtip positions in three-dimensional coordinate space.

The kinematics obtained in this study provided the opportunity to calculate force coefficients obtained from both the normal-forces model developed by Usherwood and Ellington (Usherwood and Ellington, 2002a) and thin-airfoil theory. Thin-airfoil theory resolves the resultant aerodynamic force on the wing into the components of lift and drag forces. Lift is defined as perpendicular to the direction of the airflow; drag as parallel to airflow(Norberg, 1990). By contrast,the normal-forces model uses the assumption that the resultant force on the wing is perpendicular to the wing itself(Usherwood and Ellington,2002a). This force is then resolved into components that are vertical and horizontal in the global reference frame. For both models, we estimated the instantaneous resultant aerodynamic force, F_R(t), by multiplying body mass by the vector sum of gravitational acceleration and instantaneous CoM acceleration. This estimate does not include the thrust necessary to overcome parasite drag, but estimates of parasite drag were all less than 0.5% of body weight. We therefore believe our estimates of F_R(t) to be close to the actual value. To estimate coefficients during mid-downstroke, we averaged the values of F_R(t) and the other input parameters over mid-downstroke (mds).

Angle of attack (AoA) was calculated as the angle between the wing chord at the wrist, and the resultant velocity of: the velocity at the wrist, relative to the shoulder, and the CoM velocity. Tobalske et al. included the vertical component of induced velocity when calculating AoA(Tobalske et al., 2007). We did not follow their approach, though doing so for the data presented here would decrease the calculated value of AoA by ∼10°.

Potential energy change (ΔPE, or ΔPE per cycle) was calculated from the weight of the bird and the change in CoM vertical position over the course of the wingbeat. `PE flight power'(P_PE) was calculated by dividing ΔPE by the duration of the wingbeat, Δt_wingbeat.

For statistical analysis, flight angles were grouped into categories of`–60', `–30', `0', `30' and `60'. All statistical tests were performed in Systat (Version 10.2; 2002; Systat Software, Inc., San Jose, CA,USA). Because Bird 1 did not perform –30° flights suitable for analysis, repeated-measures analyses excluded data for –30° flights. For the same reason, multiple paired t-tests had two degrees of freedom when comparing 30° flights with other conditions, and three degrees of freedom otherwise. The sequential Bonferroni method(Rice, 1989) was used to determine significance of comparisons between each pair of conditions. Linear regressions were performed on all standard wingbeats for all conditions.

Variation among individuals was not significant for any of the data presented here. In the figures, lines connect means of values at each condition, and asterisks denote significant differences, as determined by repeated-measures analyses. Values are reported as mean ± s.e.m. Representative flights (all from Bird 3) are included as Movies 1–5 in supplementary material.

Actual angles of flight were generally close to the angle of the setup(Fig. 4A, Fig. 5B). The 60° ascending flights averaged a vertical flight angle of 60±1°, and 30°ascents averaged 27±1°. Level flights showed a slight negative flight angle, –4±1°, because the birds tended to jump upward when taking off, and then descend gradually as they approached the opposite perch. The –30° descents averaged –29±1°, and–60° descents averaged –53±1°. Vertical distances traveled during the standard wingbeat were not significantly different between 30° and 60° flights (0.30±0.02 m and 0.23±0.04 m,respectively; P=0.35), nor between –30° and –60°flights (–0.35±0.04 m and –0.35±0.02 m,respectively; P=0.16). Horizontal distances traveled during the standard wingbeats for 30°, 0°, and –30° flights did not differ significantly from each other (0.48±0.05 m, 0.64±0.03 m and 0.61±0.03 m, respectively; P⩾0.019 for each comparison). Horizontal distance traveled for 60° flights was significantly less than all other conditions (0.18±0.01 m; P⩽0.007 for each comparison) and that for –60° flights was significantly less than that for 0° and –30° flights(0.26±0.01 m; P=0.001 and P=0.010, respectively).

Flight speed ranged from 2.7 to 5.0 m s^–1 and varied significantly across the four birds over the range of vertical flight angles studied (repeated-measures ANOVA, P<0.001, F=31.457; Fig. 4B). Flight speed averaged 3.6±0.2 m s^–1 for –60° flight and increased to an average of 4.9±0.1 m s^–1 for –30°flight, which showed the greatest flight speed for all conditions. Flight speed decreased with increasing flight angle to 2.9±0.1 m s^–1 for 60° ascent. The vertical component of flight velocity (`vertical flight velocity', Fig. 4B) showed the steepest increase from shallow descent to shallow ascent, increasing from –2.4±0.1 m s^–1 at–30° to 1.8±0.1 m s^–1 at 30°. Vertical flight velocity for –60° descent was –2.9±0.2 m s^–1 and was not significantly different from that at–30° (P=0.11).

Wingbeat frequency (WBF) ranged from 6.1 to 9.6 Hz over all standard wingbeats analyzed (Fig. 4C). Differences in WBF values among conditions tended toward varying significantly among flight angles (repeated-measures ANOVA, P=0.053, F=3.766) and, with a larger sample size, this may well have been borne out. WBF was more variable among birds for ascending flights. Angles of wing excursion, as measured from the shoulder to the wingtip, also did not vary significantly across flight angles (repeated-measures ANOVA, P=0.317, F=1.356). Shoulder–wingtip angles of excursion were quite large, reaching a maximum of 228° for Bird 2 during–60° flight. Angles of wing excursion measured from shoulder to wrist were more modest, but still reached a maximum of 153° (Bird 1 at 30°). The angular velocity of the wing at mid-downstroke showed a pattern similar to WBF, reflecting the similarity of wing excursion angles across flight angles.

Wingstroke paths are illustrated in Fig. 5B,C. For both steep ascent and steep descent, the wingtip always remained craniad to the wrist, indicating that the wing was held in a more protracted position. In level flight, the path of the wingtip entirely included the path of the wrist (in lateral view). During shallow ascent and shallow descent, the pattern of wingtip and wrist paths was intermediate. The kinematic pattern in the lateral view thus appears symmetric, as the wing is more protracted for steeper flights, whether ascending or descending. The paths of the wingtip and wrist in dorsal view showed little variation across flight angles, except for somewhat greater stroke amplitudes for 30°,60° and –60° flights, relative to level flight.

For ascending flight trials at 30° and 60°, the potential energy change over a full wingbeat cycle (`ΔPE per cycle', Fig. 6A) averaged 1.1±0.1 J and 1.4±0.1 J, respectively, but did not differ significantly (P=0.17). For –30° and –60° descent trials, ΔPE per cycle was again similar (P=0.32),averaging –1.6±0.1 J for both. As noted above, level flight trials were slightly negative (–0.28±0.03 J) due to the slight descent the birds used for the 0° perch setup.

Because wingbeat frequency (Fig. 4C) and hence cycle duration, did not show a strong pattern of variation across flight angles, PE flight power(Fig. 6B, P_PE) showed a pattern similar to ΔPE. Ascending flights at 30° and 60° averaged PE flight powers of 8.2±0.3 W and 11.0±0.5 W, respectively, which were significantly greater than P_PE for level flight (–1.8±0.2 W; P<0.001 for both comparisons) and significantly different from each other (P=0.002). P_PE of descending flights at –30° and –60° averaged –11.1±0.9 W and–13.0±0.5 W respectively, and were not significantly different from each other (P=0.75). Regression of PE flight power against flight angle showed a slope of 0.234 W deg.^–1, which was significantly different from zero (r²=0.95, P<0.001).

Wing stroke plane angle relative to the horizontal (SPA; Fig. 5A, Fig. 7A) did not differ significantly among flight angles. Though the difference was not significant,shallow and level flights showed somewhat steeper SPAs than did steep flights. Body angle in the global frame was steepest for 60° ascent and most horizontal at –30° flights, with –60° flight showing a mean body angle intermediate to level and shallow ascent(Fig. 5A). Body angle showed significant variation overall (repeated-measures ANOVA, P=0.004, F=9.353), but post hoc paired t-tests were all non-significant with the Bonferroni correction (P⩾0.014 for each comparison). The stroke plane angle relative to the body angle(Fig. 5A) was greatest for 60° ascent and decreased with decreasing flight angle, with a minimum at–30°.

Angle of attack at mid-downstroke (AoA) was greatest for steep descending flight (68±3°, Fig. 7B). AoA decreased as flight angle increased to –30°,0° and 30° (AoA=47±1°, 40±1° and 35±2°, respectively). AoA increased slightly as flight angle increased to 60° flight (AoA=41±3°). Differences in AoA between–60° and 0°, 30° and 60° tended toward significance(P=0.0054, 0.006 and 0.007, respectively), but following the sequential Bonferroni method, they were not deemed statistically significant.

The power calculations for the different vertical flight angles are shown in Fig. 8. Total power(P_tot) estimates ranged between –5.13 and 26.2 W and varied significantly across flight angles (repeated-measures ANOVA, P=0.001, F=15.381). Induced power (P_ind)calculations ranged between 3.5 and 24.6 W, and on average were greater for descending flights. P_ind also varied significantly across flight angles (repeated-measures ANOVA, P=0.006, F=8.450). Power due to changes in kinetic energy varied between 4.2 and –6.6 W,but did not vary significantly from zero (P=0.239) or across flight angles (repeated-measures ANOVA, P=0.438, F=0.995). Parasite power averaged less than 0.09±0.01 W for all conditions, and profile power averaged less than 0.67±0.09 W for all conditions. Because of their small magnitudes relative to P_tot, parasite and profile powers are not shown in Fig. 8A. Differences between P_tot and P_lev+P_PE were not significant for any of the flight angles (paired t-tests, P=0.134 for–60° and P⩾0.411 for all other conditions).

Across flight angles, the magnitude of F_R averaged between 4.18 and 5.20 N, and was similar to the bird's weight(P⩾0.108; Table 3). F_R magnitude was somewhat greater for descending flights. The variation in F_R across flight angles was significant overall but not among conditions (repeated-measures ANOVA, P=0.028, F=4.899; paired t-tests, P⩾0.028 for each comparison). Moreover, F_Rorientation was similar to the vertical (P⩾0.465 for all conditions) and did not vary significantly across flight angles(repeated-measures ANOVA, P=0.861, F=0.247). The force coefficients C_v, C_h and C_L did not show significant variation among conditions(repeated-measures ANOVA, C_v: P=0.137, F=2.387; C_h: P=0.156, F=2.213; C_L: P=0.159, F=2.188). Marginally significant variation for C_D was observed overall(repeated-measures ANOVA, P=0.043, F=4.106), but post hoc paired t-tests were all non-significant(P⩾0.083 for each test). The calculated values of these coefficients are shown in Table 3.

In this study, we sought to determine whether the power necessary for a bird to fly at an angle could be estimated as the simple sum of the total power required for level flight (P_lev) and the power necessary for the center of mass change in potential energy(P_PE). We found that this was generally the case. The sum of P_lev and P_PE for each angle was not significantly different from the total power (P_tot) for any ascending or descending flight angle, though the difference was greatest for –60° flight (Fig. 8B). This larger difference was due to the increased induced power requirement calculated for steep descent(Fig. 8A). That P_tot and P_lev+P_PEwere similar implies that pigeons generally do not balance the power lost or gained through potential energy change by a corresponding increase or decrease in kinetic energy (P_KE) or induced power(P_ind). Thus, as expected, ascending flight requires more power than level flight, and descending flight requires less power than level flight. As noted, however, during –60° flight the birds'deceleration and slower flight velocity led to an increased induced power requirement, resulting in a higher P_tot and an increased difference between P_tot and P_lev+P_PE for steep descent.

The average total power estimated in this study for level flight(7.96±3.09 W) was not dissimilar to that found by Pennycuick(Pennycuick, 1968a) for pigeons in level flight at corresponding speeds (∼9.5 W). However,calculation of aerodynamic power using differential pressure measurements on the wings of pigeons of similar mass(Usherwood et al., 2005) gave a value of 26 W during level flight, considerably higher than the value calculated here. Actual power output by the pigeons is likely to have been higher than the estimates of power presented here because the calculation of induced power assumes an ideal momentum jet with small tip losses(k=1.2). Calculations of induced power are thus minimum values for what may be expected in comparison to prior measurements of aerodynamic power(Usherwood et al., 2005) and muscle mechanical power (Biewener et al.,1998; Soman et al.,2005) of pigeons during flight.

The ΔPE per wingbeat cycle showed an interesting pattern between shallow and steep flights. For both ascending and descending flights,there was little difference in the ΔPE per cycle for shallow versus steep flight, suggesting that the birds were limited in the potential energy they could gain or dissipate during a single wingbeat cycle. This result, in addition to the data showing shorter horizontal distances traveled during steeper flight (Fig. 4A), indicates that in order to fly at steeper angles, the pigeons primarily regulated the forward distance traveled per wingbeat.

Despite the large range of flight angles and the large changes in PE that this required of the birds during flight, F_R was always similar to the vertical and had a magnitude similar to the bird's weight. Although perhaps counterintuitive, this result is not unexpected, as the dominant force a bird must overcome while flying at low speeds is its weight, regardless of its flight direction. Because the difference in aerodynamic force necessary to maintain steady slow flight at even extremely different angles of ascent or descent is small, the kinematic differences underlying the production of aerodynamic force are also likely to be small. Such differences, therefore, may be difficult to distinguish, even when employing the 3D kinematics methods used here to evaluate flight performance. This likely explains why we often did not observe distinct kinematic patterns that varied with flight angle.

Nevertheless, we did see trends in the data relative to flight angle. As expected, angle of attack (AoA) was highest for descending flight(Fig. 7B). But contrary to our expectations, AoA did not decrease from level to ascending flights. This indicates that for ascending flight the birds altered the angle of their wings and wing movement (stroke plane angle) in such a way that the wing angle did not change relative to the resultant angle of the CoM velocity and the wrist velocity relative to the shoulder. Overall, AoA was quite high, averaging 46° across all standard wingbeats and reaching 77° for Bird 1 at–60°. Because of the high values of AoA, the thin-airfoil model may not be applicable here, even as modified for this study, as it assumes a thin airfoil with modest AoA.

Flight speed was not constant across flight angles. Steep ascending flights were the slowest, suggesting that the energy expended by the bird was directed more toward increasing its CoM potential energy than flying more quickly. Though the differences were not statistically significant, on average the fastest flights were observed for shallow descent, which was also the flight angle for which P_tot was lowest(Fig. 8B). This suggests that the birds may take advantage of gravity to reach the lower perch. However,during the steepest descent angle (–60°) studied, the birds decelerated and flew more slowly, reflecting an attempt to maintain control along a flight path that was essentially a steeply angled fall.

Wingbeat frequency (WBF) varied, but was not significantly different across flight angles, indicating that faster wingbeats may not be necessary for inclined flight as we had expected. However, a larger sample size may have borne out a significant increase in WBF across flight conditions, with the greatest WBF occurring at the steepest flight descents and ascents(Fig. 4C).

Stroke plane angle (SPA) also varied, but not significantly across vertical flight angles. SPA was somewhat more horizontal at steep flight angles,suggesting the pigeons may have forced more air downward. For steep ascending flight, this could assist in upward propulsion; for steep descending flight,it could serve to slow the bird and give it more control. In the bumblebee Bombus terrestris (Dudley and Ellington, 1990), and three bird species(Tobalske and Dial, 1996; Tobalske et al., 2007), SPA was found to be more horizontal for slower flight, when forcing more air downward may assist with weight support. Steep ascent and descent were the conditions for which the slowest flights were observed, so the more horizontal orientation of the stroke plane may be due in part to the low flight speed.

For level flight, wingstroke kinematics(Fig. 5B,C) were similar to those depicted in Tobalske and Dial(Tobalske and Dial, 1996) for pigeons in level flight at 8 m s^–1. In level flight, the path of the wingtip, viewed laterally, surrounded the entire path of the wrist. As flight became progressively steeper, for both ascent and descent, the wingtip path was more craniad, indicating that the pigeons used a more protracted wing position throughout the wingstroke for steep vertical flight angles. Tobalske and Dial (Tobalske and Dial,1996) observed that as flight speed increased in pigeons, the wingtip path became more caudad. The kinematics seen here may reflect a continuation of this trend to lower speeds. The slowest flights observed in this study were during steep ascent and descent, which were also the conditions that showed the most craniad position of the wingtip path.

The present study examined flight at vertical angles from –60° to 60°, at 30° increments. Dial and Biewener(Dial and Biewener, 1993)studied muscle function in smaller wild-type pigeons during a variety of flight modes, including 90° vertical ascent and near-vertical(∼–80°) descent. During vertical ascent, they observed the highest wingbeat frequencies, averaging 9.1±0.3 Hz. This value fits within the range observed here for 60° ascent (6.9–9.6 Hz), but is somewhat higher than the average at this flight angle (8.3±0.6 Hz). In near-vertical descent, Dial and Biewener(Dial and Biewener, 1993) found that WBF averaged 8.8±0.5 Hz, which is above the range for–60° flight observed here (7.8–8.6 Hz). Once again, this suggests that, although the variation in WBF among flight angles was not significant here, there may be an overall trend of increasing WBF with increasing steepness of flight ascent and descent, though differences in body mass of the pigeons used in the two studies may also affect this pattern.

The data presented here for flight speed in pigeons can also be extended using data from Dial and Biewener (Dial and Biewener, 1993). For 60° ascent, average flight velocity was 2.9±0.1 m s^–1. During vertical ascent, Dial and Biewener found that the average speed was 2.6 m s^–1. This suggests that flight speed decreases further for ascent angles beyond 60°. However, this difference between 60° and 90° ascent (0.25 m s^–1) is not as great as the difference between shallow and steep ascent (1.29 m s^–1), so flight speed may begin to be constrained at an angle below 60°.

Although we did not explore the contributions of the tail to aerodynamic forces produced by the pigeons in this study, preliminary data suggest that tail spread may vary approximately 30° and tail angle relative to the body approximately 20° across the conditions we examined here. Future analysis of these contributions during different modes of flight would, therefore, be interesting to explore in more detail.

An understanding of ascending and descending flight would also be enhanced by comparing how changes in flight velocity affect the kinematics and power requirements of flight at different flight angles. Some of the kinematic observations made here for 60° ascents and descents, such as steeper stroke plane and more protracted wing position, may be due in part to the slow speeds the pigeons used for steep flight. Analysis of faster flights might clarify the role of these factors in ascending and descending flight. The use of a larger flight arena may allow birds to select preferred flight speeds for different flight angles, which may lower their aerodynamic power requirement.

The methods we used to estimate aerodynamic power requirements were based on kinematic data and aerodynamic theory, which required the estimation of force coefficients as well as assumptions about the forces acting on the wings and bird. The application of other methods for calculating aerodynamic power to the situations of ascending and descending flight, and the comparison of such results with those presented here, would help determine the validity of the methods used here. Differential pressure sensors would provide data from which the forces on the wing could be calculated(Usherwood et al., 2005). In vivo muscle data would allow measurement of mechanical power output of the flight muscles (Dial and Biewener, 1993; Biewener et al., 1998; Soman et al.,2005). These other methods may be particularly useful in understanding the aerodynamic forces and power requirements in steep descending flight, where the greatest difference between P_tot and P_lev+P_PEwas observed. Analysis of multiple wingbeats along an inclined flight path might also provide better calculations of aerodynamic power; however, this must be traded-off against kinematic resolution of individual wingbeats, which was our focus here.

Nevertheless, our results show that, once pigeons achieve a steady flight path across a large range of ascent and descent angles at relatively slow speed, the primary force that must be produced is to support their weight. It could be that the initiation of or transition to ascending or descending flight, particularly at higher speeds, would exhibit larger changes in resultant aerodynamic force magnitude and direction. An examination of how birds initiate changes in flight angle, or maneuver to fly above or below an obstacle, is likely to clarify how flight kinematics and forces are adjusted to execute these flight behaviors.

While wild rock pigeons are cliff-dwellers and likely to be well-adapted to ascending and descending flight, some species, such as aerial predators, are likely more specialized for fast ascent and descent. Determining what morphological traits make a species a faster flier at steep vertical angles,such as wing shape and sweep angle, wing loading or muscle morphology, would also likely provide further insight into the aerodynamics and biomechanics of ascending and descending flapping flight.

LIST OF SYMBOLS AND ABBREVIATIONS

 
• a_z

  vertical component of whole-body acceleration

 
• A₂

  area of both wings

 
• A_disc

  area of actuator disc

 
• A_fron

  frontal area of bird

 
• A_i

  area of ith wing section

 
• AoA

  angle of attack

 
• CoM

  center of mass

 
• c_r

  wing chord at distance r

 
• C_D

  coefficient of drag

 
• C_D,par

  coefficient of parasite drag for the body

 
• C_D,pro

  coefficient of profile drag

 
• C_h

  horizontal force coefficient

 
• C_L

  coefficient of lift

 
• C_v

  vertical force coefficient

 
• F_D

  drag force (force parallel to direction of wing movement)

 
• F_h

  horizontal component of resultant aerodynamic force

 
• F_L

  lift force (force normal to direction of wing movement)

 
• F_R

  resultant force magnitude

 
• F_v

  mass ^* vertical acceleration

 
• F_x

  mass ^* horizontal (forward) acceleration

 
• g

  gravitational acceleration

 
• i

  index for wing section

 
• I

  number of wing sections

 
• k_ind

  induced velocity correction factor

 
• m

  mass

 
• P_aero

  aerodynamic power

 
• P_ind

  induced power

 
• P_KE

  power due to change in kinetic energy

 
• P_lev

  power for level flight, excluding potential energy change

 
• P_par

  parasite power

 
• P_PE

  power due to change in potential energy

 
• P_pro

  profile power

 
• P_tot

  total power

 
• P_tot,lev

  total power for level flight

 
• r

  distance on wing from shoulder

 
• r_i

  distance of wing section from shoulder

 
• R

  length of wing

 
• SPA

  stroke plane angle

 
• t

  time

 
• v_wt

  global velocity of the wingtip

 
• V

  flight velocity

 
• V(t)

  instantaneous velocity

 
• V_x

  horizontal component of flight velocity

 
• w

  induced velocity

 
• WBF

  wingbeat frequency

 
• α′

  angle between the flight velocity and the actuator disc

 
• αdisc

  angle of the actuator disc, in the global reference frame

 
• αglobal

  angle of the wing, in the global reference frame

 
• β

  angle between direction of force and direction of wing movement

 
• δ

  angle between the stroke plane and the actuator disc

 
• ρ

  density of air

 
• ω

  angular velocity of wing

 
• ΔKE

  change in kinetic energy

 
• ΔPE

  change in potential energy

 
• Δt_wingbeat

  wingbeat duration

 
• Φ

  angle of excursion

Subscripts:

 
• lev

  level flight

 
• mds

  mid-downstroke

We would like to thank Pedro Ramirez for caring for the pigeons, and the many people at the Concord Field Station who assisted in the experiments. We greatly appreciate the expert guidance that Jim Usherwood provided in the early stages of data analysis. We are also grateful to Ty Hedrick and Bret Tobalske for their expert guidance and for helpful comments on earlier versions of this paper, as well as the critical advice offered by two referees.