THE JOURNAL OF RAPTOR RESEARCH A QUARTERLY PUBLICATION OF THE RAPTOR RESEARCH FOUNDATION, INC. VoL. 30 June 1996 No. 2 Raptor Res. 30(2):57-61 J. © 1996 The Raptor Research Foundation, Inc. ACCURACY OF ESTIMATING THE SPECIES AND SIZES OF OSPREY PREY: A TEST OE METHODS David N. Carss Institute of Terrestrial Ecology, Hill ofBrathens, Glassel, By Banchory, Kincardineshire, AB31 4BY, Scotland, U.K. Godfrey J.D. Avian Ecology Unit, Department ofBiological and Molecular Sciences, University ofStirling, Stirling, EK9 4LA, Scotland, U.K. — Abstract. The accuracies of examining uneaten prey remains collected at feeding sites and ofdirectly observing fish captured while birds forage, common methods of determining the species composition and size structure of prey in the diets of ospreys (Pandion haliaetus were tested during the summer of ), 1992 at two shallow lakes in northeastern Scotland. Prey remains were collected below feeding perches and the number ofheads and pairedjaws was used to estimate the minimal number of each species in the diet. Key cranial bones were used for species identification and length estimation. Direct field observations were also made to identify the species and sizes of fish taken by foraging ospreys. Fish species were identified by body shape and lengths were estimated by comparison with the size of the ospreys. The accuracy of field observations was tested experimentally using a life-sized model osprey and a selection ofnorthern pike {Esox lucius) and perch {Pereafluviatilis) ofvarious sizes. Results showed that the analysis of prey remains gave an accurate estimation of the size range of osprey prey, although small fish (<25 cm) were underrepresented. Tests of field observations showed that most fish could be correctly identified on the basis of their body shape but there were consistent inter-observer differences in fish length estimations. These differences should be considered in studies using field estimates of prey size, particularly those involving energetic calculations where small errors in length estimations can lead to large errors in estimations of mass and, hence, energy. Key Words: osprey, Pandion haliaetus; diet, prey estimation-, field techniques. Exactitud de la estimacion de tipos y tamanos de presas de Pandion haliaetus: una prueba de metodos — Resumen. Las exactitudes de examinar presas no comidas que permanecen en los comederos y de observaciones directas de peces capturados mientras las aves se alimentan, metodos comunes de deter- minacion de la cornposicion de especies y tamano de la estructura de presa en la dieta de Pandion haliaetus, fueron probados durante el verano de 1992 en dos lagos superficiales al noreste de Escocia. Los restos de las presas fueron colectados bajo comederos; el numero de craneos y pares mandibulares fueron usados para estimar el numero minimo de cada especie en la dieta. Claves de huesos craneales se usaron para la identificacion de especies y estimacion de longitud. Tambien se hicieron observaciones de terreno para identificar los tipos y tamanos de peces capturados por aguilas pescadoras. Las especies de peces fueron identificadas por la forma del cuerpo y la longitud fue estimada por comparacion con el tamano de la misma aguila. La exactitud de las observaciones de campo fue probada experirnental- mente usando un modelo “life-sized” del aguila pescadora y una seleccion de varios tamanos de Esox lucius y Perea fluviatilis. Los resultados mostraron que el analisis de restos de presa entregan una esti- macion exacta del rango de tamano de las presas del aguila, aunque los peces pequenos fueron sub- representados. Pruebas de observaciones de campo, mostraron que la mayoria de los peces podria ser identificado correctamente sobre la base de su forma corporal, en cambio hubo consistentes diferencias 57 . 58 Carss and Godfrey VoL. 30, No. 2 entre observadores respecto a las estimaciones del largo. Estas diferencias podrian ser consideradas en estudios usando estimaciones de campo del tamano de presa, particularmente aquellas que envuelven calculos energeticos donde pequenos errores en las estimaciones de longitud podrian llevar a cometer grandes errores en estimaciones de masa y por lo tanto de energia. [Traduccion de Ivan Lazo] The species composition and size structure of oculars or a 15-65X70 telescope. Dives were classed as osprey (Pandion haliaetus) prey have been deter- successful if a fish was seen to be carried away and un- successful ifno fish was carried. The species offish taken mined by collecting uneaten prey remains at nests was identified from its body shape and its length was es- and feeding perches, and by directly observing fish timated by comparison with the size of the ospreys. taken while ospreys forage (Poole 1989). There are The accuracy ofdirect field observations ofospreyprey potential biases associated with each method. Us- was tested at the Institute ofTerrestrial Ecology, Bancho- ing the first, the frequency of small fish may be rsyp,anusi=ng15a5licfem-)sizeanmdodaelseloescptrieoyn o(fbopdiyke=a5n5d cpme,rcwhinogf underestimated in the diet if, for example, they are various sizes that were caught in the study lochs. Fish completely ingested or their remains are hard to were suspended between the talons of the model osprey m find. Conversely, overestimates may occur if large which was then raised approximately 5 into the air for Items are removed preferentially by scavengers a period of 10—20 sec. The model was observed against the sky from a distance similar to that encountered in such as corvids or foxes. The second method may the field (ca. 150 m). Ten pike (fork lengths [FL] = 10, also be biased because field identification and size- 11, 12, 21, 21, 36, 36, 39, 40, 50 cm) and three perch estimation of fish may be inaccurate (see discus- (FL = 8, 9, 12 cm) were shovm, 10 of which were pre- sion in Carss and Brockie 1994 for osprey and also sented twice. Fish were presented in arbitrary order and observers had no prior knowledge of the range of sizes Bayer 1985, Cezilly and Wallace 1988 for other spe- to be expected. At some point during the trial, the model cies) . In this study, we tested the errors associated osprey was shown without a fish, giving a total of 24 pre- with both methods of assessing osprey diets. sentations. Six observers, including the two authors, took part in the tests for a total of 144 observations. Datawere StudyArea and Methods analyzed by linear regression ofthe relative errors in the Data on osprey prey were collected at two lakes, Loch estimated fork lengths ([estimated — actual]/actual) on Davan (42 ha) and Loch Kinord (82 ha), in the Dinnet the actual fork lengths offish presented to each observer. National Nature Reserve in northeast Scotland from We tested for differences in either the slopes or the in- tercepts ofeach observer’s estimation equation assuming June-August 1992. Pelagic fish species in these shallow (mean depth = 1.2 and 1.5 m, respectively), “kettle- (a) a different slope and a common intercept or, (b) a hole” lochs were principally northern pike (Esox Indus) different intercept and a common slope for each observ- and perch (Pereafluviatilis) The only other fish was the er. common eel (Anguilla anguilla). The northern pike is a common top predator of freshwater ecosystems in Eu- Results and Discussion rope and North America and often found in association Remains of 101 individual fish were collected be- with perch; such simple fish communities are relatively common tween June-August. The majority of remains col- in Scotland. Prey remains were collected below feeding sites (main- lected were fish heads, although some tails and in- ly telegraph poles but also trees) throughout the reserve tact carcasses were also found. Remains were most- and in adjacent areas. The number of heads or paired ly those of pike (64%) with the remainder being jaws was taken as the minimal number ofeach species in perch, as was expected given the simple fish com- the diet and key cranial bones were extracted for species identification and length estimation following Carss and munity of the lakes. In general, piscivorous fishes Brockie (1994). are seldom found in the diets of ospreys (reviewed Direct field observations of foraging ospreys were in Poole 1989). Perch and pike comprise no more made by one observer (JDG) fromJune-August 1992 and H than 16% and 37%, respectively, of the diet of Eu- all daylight hours from 0515-2230 were sampled in a ropean ospreys (Cramp and Simmons 1980). variety of weather conditions. Individual, foraging os- preys were watched from the loch shore with 8X32 bin- It was clear that ospreys took a particular size- Figure 1. (a) Length classes ofperch (N= 36) estimated from prey remains collected from feeding sites, (b) Length classes ofpike estimated from both prey remains {N = 65) and field observations {N = 36) offoraging ospreys. Data collected from Dinnet National Nature Reserve,June-August 1992. June 1996 Estimates of Osprey Prey 59 (a) Perch (n = 36) (%) Frequency Estimated length class (cm) (b) Pike prey remains (n = 65) direct observations (n = 36) (%) Frequency 11-15 16-20 21-25 26-30 31-35 36-40 41-45 46-50 Estimated length class (cm) 60 Carss and Godfrey VoL. 30, No. 2 range of prey at Dinnet. Length estimates for pike Table 1. Percentages of osprey prey length estimates ranged from 19-46 cm and those for perch were correctly and incorrectly assigned to arbitrary 5 cm size from 12-36 cm (Fig. la). These size ranges were categories by each of six observers (a-f). Observers dif- similar to those reported hy Cramp and Simmons fered in their ability to correctly categorize estimations = = P = (1980) and Poole (1989), and strikingly similar to (x^ 11.03, df 5, 0.05). those estimated using the same method in central Scotland (perch: range = 18-30 cm, — 16; pike: Observer range = 24-44 cm, N — 25) (Carss and Brockie Length Estimate a b c d e f 1994). % Correct 34.8 52.2 43.5 21.7 26.1 60.9 We observed 38 fish actually captured by forag- % Incorrect 65.2 47.8 56.5 78.3 73.9 39.1 ing ospreys. All but two, perch with estimated Total estimates 23 23 23 23 23 23 lengths of 18 and 25 cm, were pike. Length esti- mates for pike (Fig. lb) ranged from 16-44 cm (:>c = 27 cm, SE = 11, = 36). Overall, size ranges determined using this method were similar to prey identification. Nevertheless, the accuracy of those obtained using prey remains with the largest identifications may be reduced in other areas proportion of fish taken in the 26—30 cm range. where confusion could arise between similarly- Although not statistically significant (x^ test on shaped fishes such as perch and roach (Rutilus ru- numbers of fish remains and observations in <25 tilus), or pike and salmonids {Salmo spp., Oncorhyn- cm, 26-35 cm, and >36 cm size classes), small pike chus spp.). (<25 cm) were less frequendy observed in prey The regression analysis showed that there was remains than during direct field observations at significant variation among observers in the esti- ~ the lochs, and fewer large pike (>30 cm) were mation of fish sizes. Both the intercepts (F5jgs seen taken than were represented in remains col- 10.7, df — 5, P < 0.001) and slopes (F5 126 ~ 13.4, lected at nearby feeding perches. We concluded df = 5, P < 0.001) of observer regression lines that estimates of osprey diets from prey remains differed significantly. We therefore concluded that probably gave a biased picture of the lengths of such differences should be taken into account in fish taken with the proportions of small fish being studies relying on length estimates in the field. underrepresented. Most (71%) of the 138 estimates were within We have found the undigested remains of fish 20% of the true lengths with those ofone observer up to 12 cm long in the guts of larger piscivorous (JDG) being consistently within 10% of the actual fish that were partially eaten by ospreys. Therefore, lengths. Most observers estimated fish lengths with- the presence of small fish remains at nests or feed- in 3-9 cm of the actual length and one observer ing sites does not necessarily imply that hsh of this (JDG) estimated them with 2-4 cm accuracy. These size have been taken directly by ospreys; such a values would likely be the same under actual field phenomenon could explain the record of a 4 cm conditions for a similar observation distance. After fish at the nest (McLean and Byrd 1991). length estimates were assigned to arbitrary 5 cm During field tests with the model osprey, all six size classes (e.g., 6-10 cm, 11-15 cm), we were un- observers were able to correctly determine when able to improve observer accuracy and 39-78% of the osprey was not carrying a fish. Most fish the estimates were still incorrectly assigned (Table A (92.8%) in the remaining 138 experimental trials 1). further increase in the range of size classes were correcdy identihed to species (5.8% misiden- used would increase the proportions of estimated tified and 1.5% unidentified). The eight misiden- lengths correctly identified, but such results would tihed fish (4 pike and 4 perch) were the smallest be increasingly less meaningful. Therefore, it is fish used in the trials (x FL = 14 cm, SE = 2.4, recommended that observers be tested before = range 8-21 cm). Presumably, larger fish were making size estimations of osprey prey in the field. correctly identified more often because of differ- The experimental trials suggested that field ob- ences in their body shape, with pike tending to be servations of fish taken by foraging ospreys would elongate and perch deep-bodied. Observations of give an accurate estimate of the proportions of actual prey captures by ospreys usually last longer each prey species in the diet but that size estimates than 20 sec and real ospreys carry live fish which of fish would be less reliable because some observ- hold their fins erect increasing the opportunity for ers were able to estimate the lengths of fish more . June 1996 Estimates of Osprey Prey 61 accurately than others. This may have important draft of the manuscript, as did Alan Poole and Peter Mc- implications for energetic studies where prey mass, Lean. Dave Elston provided statistical advice on the anal- ysis of results from the experimental trials whileJim Par- rather than its length, is a crucial factor. Because kin gave us additional osprey observations and provided body mass varies as the cube oflength, small errors on-site accommodation for one of us (JDG) in length estimation will lead to large errors in the Literature Cited estimation of mass. We found that accurate length estimates could Bayer, R.D. 1985. Bill length of herons and egrets as an estimator of prey size. Colonial Waterbirds 8:104-109. be obtained from the collection ofprey remains at Carss, D.N. and K. Brockie. 1994. Prey remains at os- feeding sites. While this also appeared to be a valid prey nests in Tayside and Grampian, 1987-1993. Scot- technique for estimating the size range of osprey tish Birds 17:132-145. prey, it underestimated the proportion ofsmall fish Cezilly, F. andJ. Wallace. 1988. The determination of (<25 cm) taken. Nevertheless, this method was far prey captured by birds through direct field observa- less labor intensive and, hence, cheaper, than di- tions: a test ofmethod. Colonial Waterbirds 11:110-112. rect observations in determining the diets of os- Cramp, S. and K.E.L. Simmons. 1980. The birds of the preys. western Palearctic. Oxford Univ. Press, Oxford, U.K. McLean, P.K. and M.A. Byrd. 1991. The diet of Chesa- Acknowledgments peake Bay ospreys and their impact on the local fish- We would like to thank Mike Harris, Hans Kruuk, Mick ery./. Raptor Res. 25:109-112. Marquiss and Sarah Wanless for acting as observers dur- Poole, A.F. 1989. Ospreys: a natural and unnatural his- ing the direct observation trials, and Hans and Phil Ba- tory. Cambridge Univ. Press, Cambridge, NY U.S.A. con for collecting some of the prey remains in the field. Mick, Sarah and Ken Nelson commented on an earlier Received 16June 1995; accepted 11 December 1995