|
|||||||||||||||||||||||||||||||||||||||||||||||||
|
EURING Newsletter, Volume 3, July 2001 In recent decades the increasing number of birders has also produced a fast increase in the number of ringed birds reported on the basis of remote reading of both metal rings and visual marks (especially colour engraved rings). These data are of great importance to study life-histories and estimate survival rates in many bird species. However, it is often difficult to assess the rate of mistakes in such readings, and an interesting experimental approach in this respect is described here by Wojciech Kania. ERRORS
IN EXPERIMENTAL READINGS OF by Wojciech Kania ORNITHOLOGICAL
STATION GDANSK Introduction Readings from a distance of the number on metal rings with the aid of a telescope or binoculars have become quite common in the last decades. Even small rings are sometimes read in this way. J. Muiswinkel identified GDANSK KJ 45022 (digits 1.5 mm high) on a Little Stint in Tunisia. For some species the resightings represent over 90% of all recoveries. The White Stork is one of the species for which distant ring identification is particularly frequent. To make remote readings easier, special stork rings were introduced, with a short number and big characters, often repeated. Up to the end of the eighties such rings were usually high (=30 mm) and put on the ulna. After Schulz (1987) had claimed it dangerous to storks, many ringing schemes started to put rings on tibia, diminished their height and changed the design. The Polish ringing scheme considered two designs of new rings. To test their readability, an experiment was performed. The analysis of the results showed quite unexpectedly that it was not the completeness of the ring number reading but the errors made by observers that posed the main problem (Fig. 1). The experiment was repeated on different occasions and the results so far obtained presented as posters at some international meetings. Additional experiments on their participants were also organised. In this paper I focus on the most important results concerning observer errors. A wider and more detailed analysis is planned to be published elsewhere. Although considerations provided below are based mainly on remote readings of stork rings, they apply probably to all rings and other marks to a large extent. Methods and material The experiment consists in reading the rings by binoculars. Five white stork metal rings with different numbers and sometimes two colour rings were exposed simultaneously. Four of the metal rings, the ones with six character numbers (two letters and four digits) representing two designs (Fig. 2) are discussed here. The first design (Fig. 2a & b), adopted only for the experimental rings, contained triplicated numbers (as the address was planned to be printed with small characters invisible from a distance, the first letter of the ring number was designed as identifier of the ringing scheme). The numerical part of the number was repeated on both sides of the two letters in the upper row, and letters were put both before and after a set of numerals in the lower row. The complete number could be read from each direction. The second design (Fig. 2, c & d), actually used by the Polish ringing scheme Gdansk, had doubled the number also put in two rows. The first two characters of the upper row were situated just above the two last ones of the lower row. As at least four characters of one row were visible from each direction, the full number could be read without changing position of the observer from 60% of directions. All letters and digits engraved on the rings were filled with black paint. The rings were exposed on wooden sticks at various turn of the vertical slit to the observer (Fig. 2). Distances between sticks were 15-25 cm. The distance from rings to the parallel path, along which observers could move, was usually 20-22 m. The length of the path was 17-30 m. The experiments were carried out during ringing courses or ornithological meetings. The time allowed for reading was not limited. There were two experiments, with naive and then with trained observers. In the first one the participants were only told that they should read and note the number in the same manner as during observation of a ringed bird in nature. Candidates for ringers were additionally informed that errors made would be taken into account when assessing their ability to be a ringer. In total 86 people in 6 groups took part in the experiments. They were Polish or German amateur or professional ornithologists. The second experiment was performed with Polish ringing course participants who took part in the first experiment and in the discussion of the mistakes made then, where the main conclusion was that the observer should have noted down what he really saw and not what he thought was written on the ring. The circumstances of the reading as well as the design and position of the three rings were the same as during the first experiment (the fourth ring was sometimes put upside down), only ring numbers differed. Results of 25 observers (2 groups) who were later accepted as ringers are analysed here. Fisher's exact two-sided test was used to calculate significance of the differences. Results The first experiment - with naive observers In the case of a ring with a slit exposed to the observer (Fig. 2a) the most common mistake was putting the last digits before the first ones as a result of ignoring the slit (Table). It appears in 62% of the readings, whereas other errors - in 24% of the readings, similarly as in the case of the next ring (Fig. 2b). Apart from this error, 88 mistakes occurred in the readings of all four rings (Table). They appeared in 56 of 328 ring readings (17%). Up to four mistakes were made during the reading of one ring number. Erroneous reading of a character dominated (83% of 94 errors in both experiments; Table). Misreading concerned mainly the characters placed on the edge of the visible part of the ring, seen sometimes only partly and at a sharp angle. Among the misspelled characters placed in the centre was "F" instead "V". There were also errors in the ring address: "GDANSK" inscription was read as "GERMANY" (a Polish observer) and "POLAND" was read where only "ND" was visible (a German observer). In many cases parts of the number visible in the upper and lower rows were written in one line, not always being connected properly (Fig. 3a and b). When the observer tried to draw characters as he actually saw them, he often did not take care of their proper position in respect to the adjacent ones and to the edges of the ring (Fig. 3c). Particular persons read the set of rings in 5-34 minutes, on average 16 min. (The averages for groups: 13-22 min.). The second experiment - with trained observers The observers, repeating the experiment after discussion of the results of the first one, made fewer errors (Fig. 2; significances of differences between the first and the second experiments for design*positions a, b and c are: pa < 0.0001, pB = 0.01, pc = 0.18). Particular persons read the set of rings in 9-55 minutes, on average 26 min. The averages for both groups: 33 min. (readings performed by the same persons during the 1st experiment lasted on average 16 min.) and 15 min. (14 min.during the 1st experiment). Discussion Errors in the experimental ring number readings were astonishingly numerous. Is it like this in the field work? Although the observers had been instructed that they should behave as when encountering a ringed bird in the field, they obviously did not do so. Not a single person read all five metal rings (and sometimes additionally two plastic ones) during the first experiment in more than 34 minutes (i.e. on average one ring in 7 min. at the best) and some did it even in 5 minutes. It is at least several times less than usual in extremely good conditions in the field where, on the other hand, quite often many hours of observations are needed to identify the ring number of a single stork (J. Ptaszyk in lett.). This shows that the observers paid much less attention in the ring reading during the experiment than under the real field conditions. The main reasons for that could be: (1) experiments concerned the behaviour of people and not birds, which to most ornithologists is less interesting and less important, thus - subconsciously - of lesser value; (2) experiments were conducted during ringing courses or other meetings and constituted only one of many items of the agenda, not the most important one to most people. The participants were in a hurry to do other things. It was quite the opposite to the field situation, where the ring identification is usually the main purpose of a many hours activity of the observer; (3) quick completion of the task by some persons prompt others to read rings with less care - not to do it for too long. For the above reasons the level of errors in ring number readings from a distance in the field is probably lower than in the experiments reported here. The observers often wrote down not what they really had seen, but their reconstruction of the partly visible text. E.g. where "nd" was only possible to see, "Poland" was noted although, e.g., also "Finland" or "Helgoland" fit well. Most commonly wrong reconstructions concerned parts of characters placed near the edge of the visible part of the ring. Another common error of that kind was joining fragments of the number from the upper and lower rows assuming that there was a two letter and four digit number as it appears in the Polish stork rings, even when the scheme address was invisible. Also when the scheme could be identified, it has to be kept in mind that some schemes change the number of characters in the ring number, usually by adding one digit (Eggers & Fiedler 1980). A false assumption as to the number of characters can lead to a wrong reconstruction of the ring number even when enough fragments of the number are read properly, especially when the number contains two or more identical characters in the adjacent positions (Fig. 4). Similarly, ignoring the vertical slit by 2/3 observers was probably a result of an a priori belief that a number written on the ring should start with letters, not numerals, and that a sequence of four digits had to constitute a set of successive digits of the number. During the second experiment, after being acquainted with the design of the rings, only 1 of 25 people again did not see the slit. Rings with the repeated number starting and ending near the vertical slit were actually used by some schemes (e.g. Eggers & Fiedler 1980). Such a design should not generally be applied (only resightings by specially trained persons could be accepted here). Other source of errors can be disturbances in perception (dyslexia) of some observers, e.g. when "F" was noted instead of "V" (both letters denote similar sound). Using similar characters (e.g. "H" and "4"; "B", "3" and" "8") can increase the probability of wrong readings. As untrained observers, even ornithologists, have made so many errors, members of the public do probably more. In reading of the numbered neck collars on Canada Geese the accidental observers made 23 times more mistakes than the member of the project crew (Raveling et al. 1991). This raises the question if resightings done by such persons can be accepted at all. The great reduction of errors in the second experiment points out to the need for training the observers of ringed birds. As in many countries they are usually ringers, ringing courses should include the exercises in proper remote identification of ring numbers. It seems that the most efficient training might be reading of rings easy to misread and next the explanation of mechanisms leading to errors. Trainees should be instructed that instead of or prior to reconstructing the number they should draw what they really have seen, noting the position of each character in respect to the adjacent ones and to the edges of the ring (Fig. 3d), especially when the scheme address cannot be read. This can much help the ringing officer to identify the ringing scheme and the number from visible fragments or to verify the number reconstruction made by an observer. When the scheme address is not identified, also the information on the ring shape (proportion of height to diameter) and on the presence or absence of the lock and its pattern, should be added. Conclusions The results of the paper, I think, are valid not only for the stork, but for any species for which resightings are performed. 1. Analysts should bear in mind that the data they work with can be erroneous to a larger extent than is usually expected. 2. Errors in the ring numbers read from a distance are in fact probably less common than in the reported experiments, mainly because field workers put much more effort to read the number properly. 3. Ignoring the vertical slit, the cause of most errors, concerns only the rings with the number (or repeated numbers) starting and ending immediately near the slit. Such rings were used only by few schemes. 4. Training of the remote ring number readers seems to be highly desirable. 5. Real level of errors in the field remote ring number readings should be evaluated for each ring design separately, e.g. by analysing reports on rings not put on birds or put on the ones found dead before resighting them (Ebbinge et al. 1991, Raveling et al. 1991). Acknowledgements P. Chylarecki, A. Zawadzka, J. Gromadzka, K. Mokwa and M. Gromadzki made valuable comments on the manuscript. S. Blawat helped with the English. References 1) Ebbinge B.S., van Biezen J.B. & van der Voet H. 1991. Estimation of annual adult survival rates of Barnacle Geese Branta leucopsis using multiple resightings of marked individuals. Ardea 79: 73-112. 2) Eggers H. & Fiedler G. 1980. Zur Kenntnis der Storchenringe im mitteleuropäischen Raum - eine Hilfe für das Ablesen von Storchenringen. Der Falke 29: 222-227. 3) Raveling D.C., Zezulak D.S. and Silveira J.G. 1991. Accuracy in recording neck-band codes of cackling Canada geese. California Fish and Game: 205-210. 4) Schulz H. 1987. Termoregulatorisches Beinkoten des Weissstorch (Ciconia ciconia). Analyse des Verhaltens und seiner Bedeutung für Verluste bei beringten Störchen im afrikanischen Winterquarter. Die Vogelwarte 34: 107-117. Figure 1 - The problem
Table - Types and number of errors.
Figure 2. Correctness of experimental ring number readings. Ring design: a & b - experimental rings with the number visible from each direction (free area assigned to scheme address in small characters); c & d - rings used by Polish scheme. "AB1234" symbolises letters and digits of the actual ring number. Turn of the ring: part of the ring exposed to observers during the experiment. Experiments: 1st - with naive observers; 2nd - with trained observers.
Figure 3. Examples of erroneous and correct noting of one of the rings used in the experiment. Note that in the example c "VH" is above "35" instead above "51".
Figure 4. Examples of different reconstruction of the ring number from correctly read fragments of the number depending on assumptions of the length of the number. For designs g and h two fragments of the number visible from different directions are shown.
|
|||||||||||||||||||||||||||||||||||||||||||||||||
|
|||||||||||||||||||||||||||||||||||||||||||||||||
