Flight of Auks (Alcidae) and Other Northern Seabirds Compared with Southern Procellarhformes: Ornithodolite Observations

This paper records ornithodolite observations of flight speeds in 11 seabird species, comprising several different adaptive types. They are compared with earlier observations (Pennycuick, 1982b) on a set of procellariiform species, which covered a wider range of body mass, but were more homogeneous in other ways. The species in the present set can be seen as deviating from a ‘procellariiform standard’, in various ways, which affect their flight performance and style of flight.

Flight speed measurements were made between 28 June and 9 July 1986 on the island of Foula, in Shetland. Foula is an isolated island of about 12 km² projected area, centred at 60°08′ N, 2°04′ W, about 24km west of the main island of Shetland, and noted for its concentrations of breeding seabirds. A detailed account of its birds, and the status of their populations in recent years, has been compiled by Furness (1983). Most flight observations were made from the southern tip of the island, where continuous streams of birds could usually be seen flying around the South Ness, between the main breeding areas on the western cliffs and feeding areas to the east.

The body masses of Shetland birds listed in Table 2 are means for breeding adults, weighed during July over a period of years, and kindly supplied by Dr R. W. Furness (personal communication). The number in each sample is listed as n_m. The gannet mass is from means of breeding adults weighed at the Bass Rock and Ailsa Craig by Nelson (1978). The masses for South Georgia birds are means for breeding adults, published by Croxall & Prince (1980). Wing measurements were taken by the author, the number in each sample being listed as n_w. The measurements of Shetland birds were collected during a visit to Fair Isle, Shetland, in 1975, except for five great skuas measured on Foula during the present study. Those of South Georgia birds were collected on Bird Island, South Georgia in 1979-1980.

Flight speeds were measured by ornithodolite. The instrument and method of operation were essentially the same as described by Pennycuick (1982a,b, 1983). The same ‘Mariner 1’ anemometer set was used as on previous occasions, mounted on a pole 2·8 m above ground. A ‘run’ consisted of a series of timed, three-dimensional positions of the bird in space, and a speed ‘observation’ was obtained by comparing each two successive positions. The Nascom 1 computer, originally used to control the instrument, was replaced by an Epson PX8 battery-powered computer, connected via a home-made parallel interface. After each run, the bird’s groundspeed, track direction, airspeed, heading, height and vertical speed were calculated and displayed immediately on the PX8’s liquid crystal display. Two BASIC programs, incorporating machine-code routines to control the ornithodolite, were used, one for aligning and testing the instrument, and the other for acquiring and recording data. The raw ornithodolite data (not computed speeds etc.) were recorded as a tape file on the PX8’s built-in microcassette recorder. On return to the laboratory, the tape files were transferred from the PX8 to disc files on a Nascom 2 computer. All subsequent analysis was done by BASIC programs on the Nascom 2.

The species on which new flight observations are reported are listed in Table 1, with two-letter identifiers which are used in subsequent tables and graphs. This is to distinguish them from South Georgia seabird species reported on by Pennycuick (1982b), which are also listed in Table 1, with the three-letter identifiers used in the earlier publication. Some additional South Georgia species, not covered in the earlier paper, are included in Table 1. Table 2 shows morphological data for the same set of species, classified into three functional groups. Measurements of observed and calculated speeds, and other quantities derived from these, are listed for the Shetland species in Table 3.

The ‘mean flapping speed’ (V_mf in Table 3) for each species is the mean airspeed for observations in which the bird was either flapping or flap-gliding. Reynolds numbers, based on this speed, are listed in Table 3 for the body and the wing. The body Reynolds number is based on the diameter of a circle whose area is the same as the maximum cross-sectional area of the body, which is itself estimated from the mass as indicated by Pennycuick (1975). The wing Reynolds number is based on the mean chord. Body Reynolds numbers ranged from 61 000 (kittiwake) to 127 000 (gannet), and wing Reynolds numbers from 77 100 (razorbill) to 161 000 (great skua, great black-backed gull, shag).

To estimate the speeds for minimum power and maximum range (V_mp and V_mr) an estimate is needed for the drag coefficient of the body, in order to calculate its effective flat-plate area. Prior (1984) reported from wind tunnel measurements that the drag coefficients of the bodies of ducks, geese and swans declined as Reynolds number increased through the range observed here, levelling off at about 0·2 (or a little less) at Reynolds numbers above 200 000. It would appear that the value of 0·43 assumed by Pennycuick (1975) for the parasite drag coefficient of any bird, on the basis of measurements on the domestic pigeon (Columba livid) and Rueppell’s griffon vulture (Gyps rueppellii), is too high for birds with well-streamlined bodies, flying at Reynolds numbers of 60000 and above. The speeds for minimum power (V_mp) and maximum range (V_mr) were estimated by the method of Pennycuick (1975), using a value of 0·25 for the parasite drag coefficient for all species. Probably the drag coefficient should be a declining function of Reynolds number, which would have the effect of slightly decreasing the speed estimates for the birds that fly at lower Reynolds numbers, and vice versa. However, the effect would be small, and Prior’s results, as presented, would not allow such a function to be estimated with a sufficient degree of precision to justify this added complication.

For each observation, the difference between the groundspeed and the airspeed was used as a measure of ‘tailwind component’. If the groundspeed is greater than the airspeed, then there is a tailwind, and vice versa. The effect of a tailwind on airspeed was represented by calculating linear regressions for each species, in which the ordinate is airspeed, and the abscissa is (groundspeed — airspeed). Similar regressions for petrels and albatrosses (Pennycuick, 1982b) were quoted by Rayner (1985) as an illustration of the respective merits of different methods of calculating the slope of the line. Rayner concluded that the reduced major axis method, which makes no distinction between the dependent and independent variables, is more appropriate in this case than the regression method, which recognizes one variable as dependent and the other as independent. As Rayner noted, neither variable is under the control of the observer, and errors in the measured airspeed appear in both variables. However, it can be argued that the tailwind component is imposed on the bird, and is therefore the independent variable, whereas the airspeed at which the bird elects to fly represents its response, making this the dependent variable. As the point seems to be arguable, the regression calculation is retained here, primarily to render the results directly comparable with the earlier observations.

As in the petrels and albatrosses observed by Pennycuick (1982b), the slopes of the regression lines were negative for all species, that is, the birds responded to a tailwind by reducing their airspeed, and increased speed in response to a headwind. This is the expected response, as explained by Pennycuick (1978). Seven of the 11 correlation coefficients were significant at the 5 % level or beyond. The regression coefficients are listed in Table 3 as ‘wind effect’, together with an estimate of significance. The y-intercept of each regression line is listed as the ‘zero-wind speed’ (V_zw), that is, an estimate of the airspeed at which the bird would fly in zero wind.

Fig. 1 shows histograms of observed airspeeds for all species, in flapping and flap-gliding flight. The minimum power speed (V_mp), and maximum range speed (V_mr), calculated from the mass and wing span by the method of Pennycuick (1975), are listed in Table 3, and also marked on each histogram, together with the percentages of observations falling below V_mp, and above V_mr. In all species, both the mean observed airspeed (V_mf) and the zero-wind speed (V_IW) fall between the calculated values for V_mp and V_mr.

Of the three species that showed large percentages of observed speeds above the calculated maximum range speed, the puffin also showed a significant negative correlation between airspeed and vertical speed. The reason appears to be that many of the puffins passed the observation site in a shallow dive, on their way from nearby cliff nesting areas to the sea, and increased their speed accordingly. The other two species (arctic skua and kittiwake) were represented by small samples, dominated by birds flying at increased speed against headwinds. Their zero-wind speeds were much less than their mean speeds (Table 3). Several species showed large percentages of observations below the calculated minimum power speed. Of these, the great skua, great black-backed gull and fulmar were reducing speed in some observations as they made use of slope-lift along the cliff. Gannets were flying well out from the cliff, but were often seen reducing speed and looking down, presumably at fish. The shag is the only species whose speed distribution is not easily explained in terms of behaviour. It showed the highest percentage of observations below V_mp (32%), and none above V_mr. The shags were flying steadily along over the sea, and were not slope-soaring, preparing to land or looking out for prey. Their speed distribution calls for another explanation (see below).

Flap-gliding is the characteristic style of powered, cruising flight in ‘normal’ Procellariiformes - that is, albatrosses, petrels and storm-petrels, but excluding diving petrels (Diomedeidae, Procellariidae and Hydrobatidae but not Pele-canoididae). Among the species in the present sample, the gannet also regularly flapglides in cruising flight, but the others typically flap most or all of the time. A bird that flap-glides when cruising must be able to flap reasonably efficiently, at a speed pot too far from that for the best glide ratio in gliding flight. The required flapping speed cannot be too much above or below the best gliding speed, otherwise the gliding phase of the flap-glide will be unacceptably inefficient.

One might suppose that the normal Procellariiformes, whose body masses range from about 30 g to 9 kg, would choose a constant value of the lift coefficient for the gliding phase of flap-gliding flight, but this is not so. These lift coefficients were determined by Pennycuick (1982b) for seven procellariiform species, and ranged from 0·28 for Wilson’s storm petrel to 1·08 for the giant petrel. These results are plotted (together with the new value for the fulmar) as the lower line in Fig. 2. It shows that the lift coefficient varies with the 0·24 power of the body mass. The slopes of all the lines of Figs 2 and 4 were calculated by the reduced major axis method, as recommended by Rayner (1985).

The three auks (Alcidae) are the only species in the present sample that use their wings in a flapping motion for propulsion under water. An incipient form of this adaptation can be seen in underwater films of plunge diving boobies (Sulidae), which steer with the partially opened wings. However, these birds do not show the penguinlike motion seen in the auks, and their forward motion seems to be derived from the momentum of the dive, or from buoyancy when returning to the surface, rather than from flapping the wings. In the Southern Hemisphere, the diving petrels (Pelecanoididae) swim in a similar way to the auks, and are generally considered to be convergent on them. Since water is some 800 times denser than air at sea level, it is remarkable that auks and diving petrels can use the same wing, and even a similar motion, for propulsion in both media. The flapping frequency is, of course, much lower in water, and the wings are held with the wrist and elbow joints sharply flexed.

Pennycuick (1986) has represented this and other adaptations by considering them as deviations from a theoretical ‘standard seabird’. The latter is defined by a mass, wing area and wing span that fall in the centre of the allometric series of the normal Procellariiformes. The ‘standard seabird’ is like a medium-sized petrel, close to the fulmar. Variations in size from the standard generate the procellariiform series of flap-gliding seabirds. The wings of boobies are indistinguishable from the standard on this basis, but adaptation for wing swimming leads to a deviation from the procellariiform lines, in the manner shown in Fig. 3. As compared to normal Procellariiformes (or boobies) of the same mass, auks and diving petrels show reduced wing span and wing area, with little change in the aspect ratio. This process is carried to extremes in penguins and other flightless wing swimmers, in which the wing is presumably optimized for swimming. Birds that swim in this manner, but also retain the power of flight, have to settle for an intermediate degree of reduction.

It is obvious that reduction of both wing span and area in the manner of Fig. 3 will lead to an increase of both gliding and flapping speeds. Less self-evidently, it was noted by Pennycuick (1986) that the gliding speed is more strongly affected than the flapping speed. If the ‘standard’ bird (upper outline in Fig. 3) normally proceeds by flap-gliding, then the bird with the reduced wing will not be able to do so. Its flapping speed will be faster than that of the normal bird, but its gliding speed will be faster still, so that it is no longer possible both to flap and glide at the same speed. Therefore auks and diving petrels do not flap-glide, but proceed by continuous flapping. This does not mean that these birds are unable to glide. The three species of auks in the present sample do glide when slope-soaring along cliffs in strong winds. Presumably they do so at airspeeds above their normal flapping speeds. Regrettably, this conjecture could not be tested directly, as no suitable observing site was found on the cliffs, from which ornithodolite observations of gliding auks could be obtained. The difficulty was to find a site at which meaningful readings could be obtained from the anemometer. Suitable sites most probably do exist on other islands with less precipitous cliffs.

The effect can be seen indirectly by considering Cu, the lift coefficient for gliding at the minimum power speed. This is much higher in auks than in normal Procellariiformes of similar mass (Table 2; Fig. 4). The top line in Fig. 4 represents C_L1 for the three auks and two diving petrels in Table 2, for comparison with the dotted line, which is the procellariiform line transferred from Fig. 2. With a value of 1·46 for C_L1, the guillemot would be almost stalled if it attempted to glide at its minimum power speed. The mean flapping speed for guillemots was 1·37 times the estimated V_mp (Table 3), but this is evidently still too slow for flap-gliding to be practicable. Values of C_LI for the diving petrels, with body masses below 150 g, are near those for albatrosses. The latter flap-glide, whereas the diving petrels do not. Probably there are two reasons for this. First, the albatrosses are obliged by lack of muscle power to fly not much faster than their minimum power speeds, whereas the diving petrels, like the auks in the present survey, probably fly considerably faster than this. Second, the albatrosses can probably glide efficiently at higher lift coefficients than diving petrels, because of the higher Reynolds numbers at which they fly.

The seven remaining species in the Shetland sample comprise three gulls (Laridae), two skuas (Stercorariidae), the shag (Phalacrocoracidae) and the gannet (Sulidae). Their values for C_L1 are plotted on Fig. 4, together with that for the blue-eyed shag. The lower line is fitted through this group of points, excluding that for the gannet. The points for this more heterogeneous group of species show considerably more scatter about their line than do the normal Procellariiformes or the wing-swimming species. The point for the gannet falls very near the normal procellariiform line. As noted by Pennycuick (1986), gannets and boobies are indistinguishable from normal procellariiform birds on the basis of their wing span or area, and, like Procellariiformes, they flap-glide in cruising flight. Gulls, skuas and cormorants mostly flap continuously in cruising flight, but not for the same reason as the wing-swimming birds. Their C_L1 values fall below the normal procellariiform line (lower line in Fig. 4), which means that their gliding speeds are lower than their flapping speeds, rather than higher as in the wing-swimmers. When gulls, skuas and cormorants glide, they do so at speeds slower than their cruising speeds in flapping flight, rather than faster.

Birds with a low C_L1 change from flapping to gliding as they slow down, whereas those with a high C_L1, if they happen to be gliding, have to change to flapping before they can reduce speed. A gull or cormorant preparing to alight on the water, first stops flapping and glides, as it decelerates prior to touching down, whereas an auk continues flapping as it flies on to the surface. An auk preparing for a cliff landing will often approach the cliff in a shallow, fast, gliding dive. It pulls up to approach the landing ledge in a steep, decelerating climb, and as it slows down, it changes from gliding to flapping, invariably flapping vigorously just before touching down.

As noted above, the mean flapping speed observed for the shag was only 1·08 times the calculated minimum power speed, and 32 % of the observations were below V_mp. The observations were not distributed upwards towards the calculated maximum range speed, as in other species, and there were no observations over V_mr. The high estimates for V_mp and V_mr (14·2 and 23·2 ms^-1, respectively) reflect the fact that the shag’s wing span (l·04m) is very short in relation to its mass (1·81 kg). To reduce the estimate for V_mp to, say, 12 ms^-1 (which would remove the anomaly) one could argue that the induced power is much lower than assumed. On present assumptions, the wing would have to behave as though its span were about 1·35 m, which seems improbable. Alternatively, the parasite drag, or perhaps the wing profile drag, would have to be very much greater than assumed. This does not seem likely either, as such an assumption, applied to the auks, would produce an anomaly in the other direction.

The best interpretation seems to be that shags, unlike the other species, really do undertake foraging flights at speeds only a little above their minimum power speed. The observation may reflect a simple scale effect, which has been discussed elsewhere (Pennycuick, 1975). As flying animals increase in size, the power required to fly increases more steeply than the power available from the muscles. There is some value of body mass at which a bird of a particular morphological type has only just enough power to maintain its maximum range speed in level flight. As the bird is scaled up further, its muscle power becomes insufficient to reach V_mr, and the speeds available to it become restricted to a progressively narrower range about V_mp. The dimensional argument predicts the trend, but does not supply a numerical estimate for any particular combination of mass and wing span at which flight at the maximum range speed should no longer be possible. The present observations suggest that the shag is beyond this point. If this interpretation is correct it can be used as a calibration, to insert numerical values into the dimensional theory.

My thanks are due to the people of Foula for their well-informed guidance to the bird life of their island, and most especially to John and Isobel Holbourn for their hospitality, and help with practical problems. I am especially indebted to Dr R. W. Furness for his help with arrangements for the trip, and for supplying information which has been used in this paper.