Climbing Performance of Harris’ Hawks (Parabuteo Unicinctus) With Added Load: Implications for Muscle Mechanics and for Radiotracking

Two female Harris’ hawks were loaned by an aviculturist, and kept in an outdoor aviary at the Patuxent Wildlife Research Center, Maryland, from November 1986 to July 1987. Both birds had the same wing span (l-16m), and were kept on a regime of feeding and daily exercise that maintained their body mass near 920 g. The birds are referred to below by their names, Siren and Nemesis, where it is necessary to distinguish between them.

During the initial training period, each bird was fitted with a harness made of 1-cm wide woven Teflon ribbon tube (Fig. 1). A piece of Velcro fastening material measuring 3·75cm×8cm was held on the bird’s back by a neck loop passing around the furculum and a body loop encircling the body behind the wings. A longitudinal strap joined the two loops ventrally. A cloth backpack unit measuring 9·5 cm × 6·5 cm × 3·5 cm for carrying added weight was attached to the harness with Velcro. Lead fishing weights were wrapped in foam padding to prevent them from shifting in the pouch. Up to 530g could be accommodated in the backpack and, for loads up to 630g, additional weights were attached to the harness straps along the breast.

Climbing flight tests were conducted outdoors, and were recorded on videotape, using the white wall of a barn as a background. This wall was approximately 12 m long by 6 m high, and was marked out with a rectangular pattern of black crosses, made with adhesive tape, and spaced 1 m apart both horizontally and vertically. A telephone pole was erected at the right-hand end of the wall, as seen from the camera position, and was fitted with two horizontal perches, extending away from the wall, at 4 m (low perch) and 7 m (high perch) from the ground (Fig. 2). The bird took off from a movable, 1·5 m high T-perch, which could be placed at various distances from the pole. The bird’s flight path was 1·9 m from the wall if it flew directly from this perch to one of the perches on the pole. Although the barn provided shelter, the air flow near the wall in windy conditions was too gusty and inconsistent for meaningful measurement, and experiments were not undertaken if the wind speed in the open exceeded 3·5 ms⁻¹.

During each flight session, both birds carried the same amount of added mass. Each bird in turn performed between 8 and 15 climbing flights. After each flight the bird was rewarded with food (typically 8–12 g), so its mass progressively increased from flight to flight. We estimated the total all-up mass for each individual flight, by recording the total amount of food given, and interpolating on the assumption that it was given in equal portions. At lower masses, the take-off perch was placed 5 m from the pole, and the bird was called to the high perch on the pole. When mass was added and the bird could not make this steep climb, the T-perch was moved further away from the pole, 1 m at a time, until the bird could reach the high perch. If the bird could not climb to the high perch from 16 m, the T-perch was moved back to the 5 m mark, and the bird was called to the low perch. If necessary, the T-perch was moved back, again in 1 m increments, until the bird could reach the low perch.

The original video recordings were made with a Panasonic WV3250 video camera and Panasonic AG 2400 VHS recorder. The frame frequency, checked by making a recording of a stopwatch, was 29·98 Hz, against a nominal 30 Hz. The camera was mounted 18·8 m from the barn wall, with its axis horizontal, and perpendicular to the wall. The lens was set to its minimum focal length, which allowed the full width and height of the wall to be included in the picture.

Using a BASIC program on a Commodore Amiga computer, fitted with a genlock device, copies were made of the original tapes, with frame numbers superimposed. The frame-numbered tapes were recorded, and also played back for analysis, on a Panasonic PV-8000 video recorder, capable of single-frame advance. Wing beat frequency was determined, using the frame numbers, by counting the frames from the second time the wings reached the full-up position after takeoff, to the last time they reached this position before the bird initiated its landing manoeuvre. Rate of climb was determined using the Deluxe Paint II painting program on the Amiga. The video picture was projected as background on the computer screen, and advanced one frame at a time. Every three frames (0·ls), a solid black circle was placed over the image of the bird’s head. The resulting line of points was then superimposed on a grid, and printed out on a dotmatrix printer. The pitch of the grid was adjusted by counting the number of pixels, horizontally and vertically, between the images of the crosses on the barn wall, and then enlarging the grid to allow for the bird being closer to the camera than was the wall (Fig. 2). In the examples of Fig. 3, the grid has been adjusted to 0·5 m pitch at the bird’s position. The bird’s height above an arbitrary datum was read directly from the grid to a nominal precision of 0·1 m, and the rate of climb was determined as the regression coefficient of height on time, in increments of 0·ls. The first point on the climb record represented the second time the wings reached the full-up position (first time after the feet left the perch). Measurement of the rate of climb began with the second point, by which time a steady rate of climb had usually been established and continued until the bird initiated its landing manoeuvre.

The observed rate of climb was regarded as the vertical component of the bird’s airspeed, assuming that there was no consistent vertical component of wind on the climb course. In the same manner, we also calculated a regression for the horizontal speed and a ‘slant groundspeed’, which was the resultant of the horizontal and vertical components of speed. It was not possible to convert this into an airspeed, as the wind blowing along the wall of the barn was not consistent enough for meaningful measurement. The birds would not fly with a tail wind component and, in general, the airspeed may be assumed to have been greater than the observed slant groundspeed.

A 50-m horizontal flight course was set up parallel to the 1·5 m high fence of a level pasture field. Three rows of markers were spaced 2 m apart horizontally, and 0·5 m apart vertically on the fence. Two perches were placed 1·5m high at each end of the course, and the bird flew from one to the other, at a distance of 1·5 m from the line of markers. The camera was placed opposite the middle of the line of markers at a distance of 19·6 m. Using the minimum focal length, 10 m along the markers and 9 m along the flight path was included in the picture. Horizontal groundspeed was determined from frame-by-frame inspection of the video recording, in the same manner as described above. In the course of an experimental session, the bird flew alternately back and forth along the flight course. The windspeed and airspeed were assumed to remain constant from flight to flight, and the airspeed was estimated as the mean of the observed groundspeeds, going left and right.

The objective of the climbing experiments was to determine the maximum rate of climb that the bird could achieve, as a function of all-up mass. We therefore excluded flights in which the bird flew to the lower perch, if it succeeded in reaching the upper perch on another flight in the same session. We also excluded flights in which the bird took off outside the field of view, since most records of this type showed the bird entering the field of view at a substantial speed, and then pulling up, as in Fig. 3B. The resulting subset of the data consisted of those flights in which the bird established a steady rate of climb within two wing beats of takeoff, and flew to the highest perch that it could reach. 65 flights by Siren and 56 by Nemesis satisfied these criteria, giving a data set of 121 flights. All-up mass varied from 906 to 1358g during these flights for Siren, and from 917 to 1347 g for Nemesis. The mean number of observation points per flight was 16·3 and the mean time between first and last points was 1·53 s.

Flapping frequency showed a small increase with all-up mass in both birds. The slopes of the two linear regression lines in Fig. 4 are significantly different from each other, and from zero at the 1% level according to a t-test. However, the lines cross within the figure. At the point where they cross, the estimates from the two regressions are of course equal, and the lines separate only slightly before they reach the left and right boundaries of the figure. The mean and standard deviation wing beat frequencies were 5·96 ±0·20Hz for Siren and 5·91 ±0·23Hz for Nemesis. The linear regression for both birds combined had a slope of 0·782 Hz kg⁻¹ and a correlation coefficient of 0·369 for 121 points.

Slant groundspeed is defined as the resultant of the horizontal and vertical speeds measured from the grid (above). Slant groundspeeds were not significantly correlated with all-up mass in either bird at the 5% level (Fig. 5). The variability of slant groundspeed is to be expected, since the windspeed could not be determined for individual flights. As noted above, winds were either light or with a headwind component, so the slant groundspeed would be an underestimate of the bird’s airspeed. However, the mean slant groundspeed for both birds (4·13ms⁻¹) was very low in comparison with the estimated minimum power speed (see Table 1) and the observed speeds in horizontal flight (below). Fig. 5 does not supply any grounds for believing that the birds increased their airspeed as the all-up mass increased (as might have been expected), and also indicates that the birds selected airspeeds on the low side of their minimum power speeds.

Linear regressions of rate of climb against all-up mass also yielded lines for the two birds with a small but significant difference in slope (Fig. 6). As with wing beat frequency, the lines crossed within the figure, and there were no significant differences of estimate, within the limits of mass that we used. We therefore combined the data from both birds to produce the graph of climbing power versus mass (Fig. 7), which is the basis of our discussion. Climbing power was calculated from the rate of climb and the all-up mass, according to equation 1. The regression equation for climbing power against mass was:

Three sets of observations, totalling 29 flights, were obtained in which the bird repeatedly flew back and forth along the course. The airspeed estimates were 10·5, 10·0 and 10·3 ms⁻¹ for mean all-up masses of 0·955, 0·978 and 1·31 kg, respectively. These speeds are close to the estimated minimum power speed, which varies from 9·9 to 11·1 ms⁻¹ over the same range. Probably the birds made these short flights at or near their minimum power speed. Videler et al. (1988) found a similar result in a kestrel carrying added mass.

Both birds flew the length of the course with their all-up mass at 1·55 kg, but they could not climb to the perch at the end with such a heavy load. We were not able to observe repeated flights in opposite directions with very heavy loads.

The flight times, on which the data in Fig. 7 are based, were all less than 2 s. Any implications for muscle power output refer to ‘sprint’ performance, that is, brief bursts of anaerobic activity. Also, we are concerned only with mechanical power actually generated by the flight muscles, not with secondary components of power, such as that exerted by the heart, or basal metabolism. When using the methods of Pennycuick (1975) to estimate the power requirements, these components have to be excluded. We used the version of the power calculation published as ‘Program 1A’ by Pennycuick (1989).

Table 1 shows two values of the climbing power, estimated from the regression equation 2, at m = 1·0 and 1·3 kg. The value of K of course changes, depending on the value assumed for b, and so do its dimensions. The estimate for power available from the muscles comes out to be 46·0 W if the aerodynamic power is assumed to vary with the 1·5 power of the mass, and slightly less (40·7 W) if P_ae varies with the square of the mass. Also listed in Table 1 is the estimated minimum power (mechanical components of power only), which is close to the values calculated for the aerodynamic power if b = 2. However, the corresponding minimum power speeds (10·0 and 11·1 m s⁻¹) are too high to be readily reconciled with the slant groundspeeds that we observed (Fig. 5). We prefer the higher power estimates corresponding to b = 1·5, as these correspond to speeds nearer to the observed slant groundspeeds (4·4 and 5·lms⁻¹ for m = 1·0 and 1·3kg, respectively).

Our results therefore suggest that the flight muscles were capable of producing 46 W in a short burst of exertion, and no less than 41W. At the mean wing beat frequency of 5·94Hz, a power output of 46·0 W corresponds to 7·74 J of work per contraction. If we knew the mass of the flight muscles, we could express this as specific work. The most closely related species for which we have good data on the mass of flight muscles is Cooper’s hawk (Accipiter coopéra), in which Marsh & Storer (1981) found that the mass of the pectoralis muscles for both sides averaged 17% of the body mass. On this basis, if we take the body mass of our Harris’ hawks as 0·920kg, the mass of their pectoralis muscles would be 0·156 kg, and the specific work would be 49·6 J kg⁻¹.

Incidental sightings of birds carrying apparently heavy prey loads have been reported in the literature (Henry, 1939; Imler & Kalmbach, 1955; Ingold & Ingold, 1987), but there is no information as to the maximum loads that particular birds can lift. The power curve calculation embodied in Program 1A of Pennycuick (1989) was used to estimate at what value of the all-up mass the minimum mechanical power would be equal to the power available from the muscles. This should be the maximum mass at which the bird is just able to fly horizontally, with no power left over for climb. Taking the estimates of power available from Table 1, the maximum mass would be 2·02kg if P_av = 40·7W and 2·18kg if P_av = 46·0 W. This would correspond to added loads of 1100 and 1260g, respectively. In fact, neither bird flew successfully with a load exceeding 625 g.

Several reasons may contribute to this discrepancy. First, the horizontal flights were of longer duration than the climbing flights, and the maximum sustainable power output may therefore have been lower. Second, the experimental conditions required the birds to accelerate to their flying speed without much loss of height. The calculated minimum power with 625 g of added load is 26·6 W, at a minimum power speed of 11·8ms⁻¹. A free drop of about 8m would allow the birds to accelerate to a flying speed of 12 ms⁻¹ with little effort, and it is possible that under these conditions they might be able to keep going horizontally with heavier loads. Another problem was that, with the heavier loads, the birds’ legs appeared to be under excessive stress, and they had difficulty in standing and balancing. Raptors are adapted to carry loads partly in the crop and partly suspended from the talons. It is possible that a raptor trained to lift a load in a more natural way, by seizing it in its talons, could lift more than our birds with their harnesses. Carrying the load in the talons would also allow the bird some control over the position of the centre of gravity, which is not the case with a harness-mounted load and this, too, may have had an adverse effect on flying ability.

For a raptor feeding nestlings, the most important aspect of flight performance is likely to be the mass of food that it can lift and deliver to the nest. In contrast to the case of a migrating goose, considered by Obrecht et al. (1988), the drag of the radio is here of minor importance compared with its mass. Our results show that our Harris’ hawks were able to take off and climb with added mass up to about 400 g, and it is reasonable to assume that they would be able to deliver food loads of this amount if they were nesting. If the bird were laden with, say, a 40-g radio, its takeoff and climb performance would be the same as before, when carrying 360 g of food instead of 400 g. Unless the bird could compensate by making more frequent kills, one might anticipate a 10% reduction in the rate of food delivery to the young, with a consequent effect on their growth rate and prospects of successful fledging. As a general approach to setting criteria for acceptable radio mass for nesting raptors, and some other birds that carry large food loads to their nestlings, we suggest that it is not appropriate to express the mass of the radio as a percentage of the body mass, as is usually done. Instead, one should first estimate the mass of food that the bird is able to lift in a typical foraging flight (from empirical field data), and then express the radio mass as a fraction of the food mass.

We thank M. Moreland for lending us the hawks, and D. S. Chu, H. H. Obrecht III and K. Titus for assisting with hawk training and handling during the experiments. R. L. Jachowski, D. G. Jorde and H. H. Obrecht III made many helpful comments on the manuscript. We appreciate the support of W. S. Seegar and F. P. Ward for this project, which was funded under Cooperative Agreement no. 14-16-0009-86-965 between the US Fish and Wildlife Service and Sea and Sky Foundation. Use of trade names does not imply endorsement by the US government.