RT11: Weighing and observing birds: Automatic data recording and processing

¹Population Biology Group, Department of Genetics, Eotvos University, Muzeum krt. 4/A, 1088, Budapest, Hungary, e-mail tothz@falco.geobio.elte.hu; ²Edward Grey Institute of Field Ornithology, Department of Zoology, University of Oxford, South Parks Road, OX1 3PS, Oxford, UK.

Pásztor, L., Perrins, C.M. & Tóth, Z. 1999. Weighing and observing birds: Automatic data recording and processing. In: Adams, N.J. & Slotow, R.H. (eds) Proc. 22 Int. Ornithol. Congr., Durban: 3182-3184. Johannesburg: BirdLife South Africa.

We discussed techniques for the automated observation of (free living) birds that allow behavioural and ecological parameters to be measured. The method in focus was observation by weighing. We were also interested to see the special potentials of the combination of different observation techniques.

In the last decades there were a few dozens of studies that involved weighing individuals at nests or on perches to collect data on body weight changes and visiting activities of a number of different bird species. The weighed object, which is the place visited regularly by the observed animals, may be a nest (box), an artificial perch or feeder. Consequently researchers can use the method to investigate bird species that nest in boxes (Great Tit Parus major, Crimson Rosella Platycercus elegans), on platforms (e.g. on buildings; Blackbird Turdus merula, Black Redstart Phoenicurus ochruros), in burrows (Sand Martin Riparia riparia), on the ground (Lapwing Vanellus, Antarctic Petrel Thalassoica antarctica); or those frequenting a perch (Red-backed Shrike Lanius collurio); or those accepting food provided by a feeder (European Starling Sturnus vulgaris). Moreover, it was possible to measure birds that are visiting on foot: penguins at the ‘gate’ to their colony or kiwis at the entrance to their nest burrow. The activity during which weights are measured may be nest building, egg laying, incubation, courtship feeding, brooding, nestling feeding, fledging; or singing, foraging, resting at a perch; or the feeding intensity and the dominance relations at a feeder.

Computers and electronic balances offered a higher level of automation for these measurements. Data may be recorded by a local computer (e.g. palmtop) beside each balance, or by one central computer connected to the balances by long cables or by radio transmission. The central computer provides effective supervisory control over the observations.

However, a relatively complete software package has been developed only recently, aiming to support not only automated, non-stop, long-run data collection but also to reduce the tremendous work required by the processing of the resulting floods of data which is needed to extract variables for statistical analyses. Zoltán Tóth developed ‘The Wisitor’ software for weighing nests and he collected data on breeding attempts in ten bird species, including tits, sparrows, starlings and blackbirds. Parts of the software were used in Oxford (Richard Woodburn, Robin McCleery), Canberra (Elsie Krebs) and Nyíregyháza (Tibor Szép).

Data recording in ‘The Wisitor’ weighing system is done on the basis of event recognition. Two series of weighing results are stored (pre-event and post-event series) at each arrival and departure. Data processing modules estimate the weight levels and their measurement error from the recorded raw data and build the visit records. The high-level support of data processing includes automated calculations with adjustable parameters, interactive graphs, automated identification of the two parents from the body weights of the visitors, and several output files with different data structures. Visit records contain identification (male/female), time of arrival and of departure, length of in and out bouts, body weight, load weight, delta parent weight.

A crucial feature of this measuring system is the potential of estimating the measurement error from the recorded weight series. This addresses a generally overlooked problem: Weighing moving objects under field conditions imposes a significant error on the measurements of small quantities. In order to detect differences between those (e.g. the average load sizes of individuals, nests, territories) it is essential to find and exclude the bad measurements and in turn reduce the average measurement error of the remaining data set. Tóth and Pásztor, with the help of Evert Meelis (Leiden), created a numeric method to estimate a major component of the measurement error for each weighing series. Filtering the data with an error limit yields more reliable estimations of both load weights and body-weight changes between consecutive visits.

These weight variables can be used to define rough behavioural acts as foraging trip with or without self-feeding, which form the basis of an analysis of parental foraging behaviour. For instance, the minimum amount of the collected food and its allocation between nestling- and self-feeding can be assessed. The time budgets of the observed birds can be reconstructed from the acts for entire nesting periods. However, without additional measurements, the energy budgets of the parent birds cannot be reconstructed from continuous weighing, as it was pointed out by Sally Ward (Aberdeen).

Ruedi Nager (Glasgow) warned that body composition of birds may change with environmental conditions without being reflected in changes in body mass. Gulls which were experimentally manipulated to lay one additional egg lost a small, but significant amount of protein from their flight muscles. This effect is small compared to the natural variation in mass, but it had very important consequences on the bird's ability to raise young.

Visual observation, the most traditional direct observation method, can provide information on the behaviour of visitors beyond visiting activity, and what’s more also on the behaviour of resident individuals, like nestlings, including interactions. When some load is involved in the observed activity (e.g. feeding) it can give information on the volume (length, width) and the identity of the cargo (e.g. classification of prey). Direct observation, however, may be too time-consuming or practically unrealisable, especially when it’s used long term, for example supplementing some automated observation. Taking pictures in an automated fashion can overcome not only these problems but it can increase the reliability and the resolution of the visually obtained data. Pictures may allow subtle identification of individuals and items (e.g. prey) and measuring the exact size of those. The drawback is that pictures are restricted in many ways (angle, focus) compared to direct observation.

Robin McCleery told about the Oxford study on Great and Blue Tits Parus caeruleus that included monitoring nests both by automated weighing (‘The Wisitor’) and by automated photography, simultaneously. The head and beak load of the entering parent was photographed at each visit, using a Super-8 mm cine camera with single-frame film-advance and flash synchronisation, triggered by a micro-switch at the entrance hole. The sex of parent, the number and type of prey delivered (often to species level) and the time were determined. The records allowed detailed analyses of the diets of nestlings between day 4 and fledging. Sample size for feeding activity data was increased by nest visit counters that were triggered by the depression of a micro-switch wire each time a bird passed through the entrance hole of the nest box.

Marcel Lambrechts (Montpellier) used video cameras and metal detectors to obtain information about the feeding frequencies of tit parents. The time of the visit, the volume and length of the prey, and its type (caterpillars, spiders, grasshoppers) were determined from the video, and information was obtained on the behaviour of the parents and the young in the nest.

A disadvantage of picturing systems is that coding and quantifying the available information into data from the records remains ‘manual’ and extremely time-consuming. Using modern digital techniques it would be possible to record pictures directly in electronic formats. Digital picture processing on a computer with a specialised software can make extracting the data more efficient by automatic selection of relevant frames, and by providing tools for measuring the size of (prey) items and maybe even for their identification.

The combination of the automatic recording of brood and parent weights with radio-tracking of the parents' range use offers some attractive opportunities. Balances weighing artificial nests made of non-moistening material quantify brood growth as well as changes in parent condition with high temporal resolution. Simultaneous records of the parents' range use therefore allow the amount of food obtained from different habitat types to be quantified. In a pilot study of the Swiss Ornithological Institute, reported by Beat Naef-Daenzer, the simultaneous application of both techniques was tested successfully on the Barn Swallow Hirundo rustica. Initial data on four pairs have demonstrated that e.g. adverse weather greatly reduces the amount of food given to the nestlings, but this is not due to a reduction in searching effort of the parents. Since swallows forage in distinct areas for about 0.5-3 hours foraging success and the amount of food taken to the nest can be quantified with a resolution of a few hours.

Elisabeth A. Schreiber (Washington, DC) talked about observations on the time and energy budgets of breeding boobies and tropicbirds. Presence and absence of radio transmittered birds was recorded by a computer to track adults' presence at the nest in order to determine feeding rates and resting times at nest. Watch activity recorders fixed on their leg were used to determine time on the water (resting) when away from the nest. Results did not support the energy limitation hypothesis, as adults spent most of their time resting and could increase feeding activity significantly when manipulated.

A multisensor telemetry system for studying flight biology and energetics of free-flying Eurasian Griffons Gyps fulvus were presented by Ralf Boegel (Berchtesgaden). The method is suitable for long-term monitoring of heart rate, body temperature, plumage temperature and air pressure (flight altitude) of large birds. For establishing good correlations between measured parameters a calibration in two ways is essential: (1) Correlating energy turnover rate versus heart rate and body temperature in a metabolic chamber (laboratory); (2) later on in the field making visual observations to identify certain patterns of measured parameters, which correlate with a certain ethological context (e.g. a fast decrease in plumage temperature is associated with flight activities, even if the air pressure sensor indicates no changes in altitude, i.e. constant level flight). After these calibrations, most of the data can be interpreted without having visual observations.

While automated measurements can produce high-resolution and accurate data - otherwise unavailable, the equipment is often expensive and without good software support data processing may become overwhelming. Constraints on sample size (of individuals) call for efficient experimental designs and a combination of direct observations with joint application of complementary automated techniques.