miércoles, 31 de agosto de 2016

Estudio de la influencia del color de las anillas de las Gaviotas patiamarillas (Larus michahellis lusitanius) respecto a la lectura a distancia con telescopio

Este estudio se realizó porque sospechábamos que la influencia del color de las anillas ( Entre las anillas rojas de Gipuzkoa con el dígito blanco y las naranjas de Bizkaia con el negro) de las Gaviotas patiamarillas respecto a la lectura a distancia con telescopio, podría tener sesgos a la hora de estudiar diferentes dinámicas de  población. (Tales como la gestión de los recursos biológicos, el tamaño, estructura de edad, sexo etc así como los factores que causan todos estos cambios). Sin embargo el estudio concluye que la  influencia del color de las anillas (Rojas con el díjito blanco / Naranjas con el dígito negro) de las Gaviotas patiamarillas respecto a la lectura a distancia con telescopio no es tal. De modo que el número de observaciones en relación con el numero de las gaviotas anilladas no varió de acuerdo con el tipo de anilla.  Ante la imposibilidad de adjuntar el artículo completo al blog he optado por hacer un copia pega que podéis leer mas abajo, en la misma entrada.

Anilla roja de Gipuzkoa junto a la naranja de Bizkaia

 Leyendo con el telescopio las anillas de colores 

Gaviotas patiamarillas con anillas naranjas de Bizkaia

 Leyendo con el telescopio las anillas de colores 

Gaviotas patiamarillas con anillas rojas de Gipuzkoa

Journal of Ornithology

Assessing the impact of colour-ring codes on parameter estimates from Cormack–Jolly–Seber models: a test with the Yellow-legged Gull (Larus michahellis)

Aldara Fernandez • Asier Aldalur • Alfredo Herrero • Aitor Galarza • Jon Hidalgo • Juan Arizaga

Received: 23 December 2015/Revised: 7 April 2016/Accepted: 25 July 2016 Dt. Ornithologen-Gesellschaft e.V. 2016

Abstract The type of ring (i.e. its colour combination: the colour of the alphanumeric characters and the base colour of the ring) used in population dynamics studies could have an effect on parameters such as resighting probability (p); for instance, we might expect a lower p value when using rings with white characters than when using rings with dark characters. We performed an experiment using two types of colour rings (white or black characters on a dark base) in Yellowlegged Gulls in order to test our hypothesis. To do this, we used Cormack–Jolly–Seber (CJS) models incorporating variation in p among the birds. We obtained no evidence supporting an effect of type of ring on either survival or p.

Keywords Birdwatching Colour ring Cormack–Jolly– Seber models Ringing MARK software


Der Einfluss von Farbring-Codes auf Parameterscha ¨tzungen mittels Cormack-Jolly-Seber Modellen: ein Experiment mit Mittelmeermo ¨wen (Larus michahellis)
Der Ringtyp (Farbringkombinationen aus alphanumerischen Kodierungen und der Grundfarbe des Ringes), derin Studien zur Populationsdynamik benutzt wird, ko ¨nnte einen Einfluss auf Bestimmungsgro ¨ßen wie die Wiederfundwahrscheinlichkeit (p) haben. Deshalb wu ¨rden wir einen geringeren p-Wert fu ¨r Ringe mit weißen Buchstaben erwarten als fu ¨r Ringe mit dunklen Buchstaben. Wir fu ¨hrten ein Experiment an Mittelmeermo ¨wen durch, bei dem wir zwei verschiedene Typen von Farbringen (weiße und schwarzen Buchstaben auf einer dunklen Grundfarbe benutzten, um unsere Hypothese zu testen. Dafu ¨r wendeten wir Cormack–Jolly–Sebel (CJS) Modelle mit verschiedenen p-Werten an. Wir ko ¨nnen keinen Nachweis liefern fu ¨r einen Effekt des Ringtyps weder auf die U ¨berlebensrate noch auf die Wiederfundwahrscheinlichkeit.

Bird individualization through ringing or other types of marks that are designed to be read at distance is one of the most commonly used tools in avian research. For instance, bird ringing is crucial in population dynamics studies where capture–recapture models are applied (Siriwardena et al. 1998; Peach et al. 1999; Aradis et al.2008; Robinson et al.2008). The generalization of the use of colour rings or other marks to be read at distance has improved our capacity to estimate parameters associated with population dynamics (Lebreton et al. 1992; Pradel et al. 1997; Papadatou et al. 2012). A notable portion of the data used in models of population dynamics is based on data reported by birdwatchers (e.g. Tavecchia et al. 2001; Juez et al. 2015). It is important to know whether some types of marks are easier to read than others in order to optimize the efficiency of field work. It is a frequent claim among birdwatchers that dark rings with clear characters are more difficult to read than clear rings with dark characters (pers. comm.). This may be due to a lower contrast between the colour of the ring and the colour of the alphanumeric characters, or the fact that clear characters often become filled with dirt, making them really hard to read at distance. The effect of the type of ring (i.e. its colour combination: the colour of its alphanumeric characters and the base colour of the ring) on parameter estimates of capture–recapture models has, to our knowledge, never been tested. In theory, estimates of survival should not be affected by the particular colour combination used in a ring, only the estimated resighting probability, p (Lebreton et al. 1992). However, according to experienced birdwatchers, a lower p value can be expected when using rings with white characters than when using rings with dark characters. Thus, the type of ring used could affect survival estimation (or other parameters which are estimated) if the model fitted assumes that p does not vary among different types of rings. Therefore, in this paper, we report an experiment we performed using two types (colour combinations) of colour rings in order to test for an effect of ring type on the parameters obtained from Cormack–Jolly–Seber (CJS) models. If the type of ring has an effect, we expected that it would influence the estimated p value but not survival.

Materials and methods

Model species
The Yellow-legged Gull is the most abundant gull of the southwestern Palaearctic (Olsen and Larson 2004). It is distributed across the circum-Mediterranean region, Iberia and the Macaronesian islands (Bermejo and Mourin ˜o 2003; Olsen and Larson 2004). Moreover, the species has also colonized some regions along the coast of France, the English Channel and some inland wetlands in central Europe (Geroudet 1984; Ye ´sou 1991; Olsen and Larson 2004). Within Iberia, the Atlantic coast hosts a resident population of more than 80,000 breeding pairs (Molina 2009), aside from the gulls of mostly Mediterranean origin that overwinter within this region (Galarza et al. 2012).

Study area and data collection
Gulls were ringed as chicks in three colonies situated along the coast of the southeastern Bay of Biscay, within the Basque region, in northern Iberia (from east to west): Ulia (43.34 N, 01.96 W), Lekeitio (43.36 N, 02.50 W) and P. Lucero (43.36 N, 03.11 W). These colonies host, respectively, ca. 500, 1200 and 50 breeding pairs (Arizaga et al. 2009).
We ringed 198 chicks in 2012 when they were ca. 20–25 days old with a metallic, engraved colour ring (Cantos 2000). In each colony, we tried to ring equal numbers of chicks with the two types of rings used in our experiment: (1) an orange ring with four dark (black) characters, and (2) a red ring with four clear (white) characters. The two types of rings had characters of a similar form and size (font size 10 9 5 mm; font width 1 mm). We used both numbers and letters to create unique character combinations. Ringing was carried out in just one or a few days at each colony, so the protocol was designed to enable the ringing of as many chicks as possible during the visits. All chicks were ringed in late June and the beginning of July and only under good meteorological conditions in order to prevent possible handling effects on nestling survival. Sighting data on alive gulls, reported by birdwatchers or by us, were compiled from August 2012 to June 2014. Sighting data were obtained from wherever these birds were seen. Overall, we compiled data on 569 sightings relating to 150 individuals (Table 1).

Capture–recapture models
We used Cormack–Jolly–Serber (CJS) models to estimate survival, which allowed us to separately estimate survival (/, the probability that a bird survives during the period from t to t ? 1) and p (the probability that a bird still alive at t ? 1 is seen at t ? 1) (Lebreton et al. 1992). We sampled monthly for 2 years, so we obtained a matrix of 24 columns (12 months from June/July of a ringing year to June of the next one) by 198 rows (individuals). Each individual gull was also assigned to one of six factors corresponding to the six colony and ring-type combinations (3 colonies 9 2 ring types). Thus, we considered six groups.

Table 1 The number of Yellow-legged Gulls ringed (as chicks) with each type of ring at each colony, the number of gulls seen alive during the subsequent 24 months, and the number of sightings obtained overall (many gulls were seen more than once)

Colony/ring Ringed/Seen (alive)/Sightings

Ulia White letters 50 31 (62.0 %) 164 Black letters 48 32 (66.7 %) 209 
Lekeitio White letters 31 18 (58.1 %) 57 Black letters 30 14 (46.7 %) 62
P. Lucero White letters 19 13 (68.4 %) 40 Black letters 20 16 (80.0 %) 37

Before starting to select models, we explored the fit of the data to CJS assumptions using a goodness-of-fit (GOF) test (Choquet et al. 2009). CJS models assume that all marked individuals have the same probability of recapture and survival from time t to t ? 1 (i.e. no trap dependence, no transients). The U-CARE software (Choquet et al. 2009) was used to run the GOF global test, which did not indicate significance (v2 = 82.009, p = 0.999, df = 142). The specific test to detect transients could not be calculated as all of the birds were marked within a single time unit (the first one in the series, corresponding to June/July of 2012). We detected a trap happiness effect (Z =-3.537, p\0.001), indicating unequal detectability among individual gulls. This is likely to be caused by the fact that the sampling effort was not uniformly spatially distributed, so gulls visiting sites associated with a higher sampling effort were more likely to be detected (i.e. they provided a higher number of sightings). CJS models were then run in software MARK 6.2 (White and Burnham 1999). Regarding /, we tried to reduce the number of models to be tested give that we were interested in models in p. Hence, we only considered combinations that assumed effects of ring type and age. The latter effect was tested due to the fact that survival during the first weeks after the ringing date (/1) is expected to be lower than the month-tomonth survival later on (/2) (Juez et al. 2015). Models assuming age dependence (/1, /2) were nested within [/ (time)], as they could be created by setting /2 parameters equal. Within our data set, where all gulls belonged to a same cohort (all birds were ringed as chicks in 2012), the effects of age and time could not be distinguished. However, even in this case, models assuming two age categories (/1, /2) fit the data better than models with time dependence (Juez et al. 2015). For p, we considered all possible single or additive combinations assuming an effect of ring type or time, both with and without the effects of mixtures of p values. We used CJS models with mixtures of p values
to account for the effects of interindividual variations in detectability (not all birds are seen with the same probability). These models included a mixture parameter (pi) in order to account for two groups of birds with a lower and a higher p, respectively (Pledger et al. 2003). Overall, we tested 16 models. We used Akaike values corrected for small sample sizes (AICc) to rank how well the models fitted to the data (Burnham and Anderson 1998). Models with an AICc difference of \2 were considered to fit the data equally well, and those for which the difference was [2 were considered to fit to the data less well. CJS models were run in the software MARK 6.2 (White and Burnham 1999), using the sin-link function.

Overall, a high proportion (62.6 %) of the 198 Yellowlegged Gulls ringed as chicks were seen during the subsequent 24 months. The ring type did not influence the proportion of the ringed gulls that were seen later (v2 = 0.914, p = 0.386, df = 1). Similarly, the number of sightings in relation to the number of gulls ringed did not vary according to the type of ring (v2 = 1.267, p = 0.283, df = 1). One model was observed to fit the data better than the rest (Table 2). This included effects of age on survival and mixture on p, which also was time dependent (i.e. p varied from month to month). Accordingly, we did not obtain any evidence supporting an effect of ring type on either / or p. Models assuming an effect of ring type on p were ranked lower than those that only assumed a dependence of p on time. Thus, the model [/1, /2, pm(ring); pm = p with mixtures] showed an AICc of 2352.78, while the model [/1, /2, pm(time)] showed an AICc of 2290.33. A third model assuming an additive effect of both factors on p [/1, /2, pm(ring ? time)] was also ranked lower (AICc:

Table 2 Ranking the best ten models used to test for the effects of several factors on survival (/: /1 is the survival between the ringing date at the end of June/July 2012 and August 2012; /2 is the constant monthly survival for the subsequent months) and resighting with (pm) or without (p) mixture effects

Models     AICc    DAICc    AICc    weight np    Deviance

1. /1, /2, pm(time)       2290.33-0.00-0.91-49-1209.71

2. /1(ring), /2(ring), pm(time)     2295.07 4.74 0.09 51 1209.68 

3. /1, /2, p(time)      2310.18 19.84 0.00 25 1284.21 

4. /1(ring), /2(ring), p(time)    2314.51 24.18 0.00 27 1284.17 

5. /1, /2, p(ring ? time)      2330.88 40.55 0.00 48 1252.64 

6. /1(ring), /2(ring), p(ring ? time)     2335.15 44.81 0.00 50 1252.15 

7. /1, /2, pm      2351.47 61.13 0.00 5 1367.67 

8. /1, /2, pm(ring)      2352.78 62.45 0.00 7 1364.90 

9. /1(ring), /2(ring), pm     2335.52 65.19 0.00 7 1367.64 

10. /1(ring), /2(ring), pm(ring)     2356.38 66.04 0.00 9 1364.38

AICc Akaike values corrected for small sample sizes, DAICc difference in AICc values in relation to the first model, np number of parameters, time time dependence (i.e. survival varies depending on the month)2380.54) than the time-dependent model. Models with no mixtures were still ranked lower than those with mixtures. Thus, the model [/1, /2, p(ring)] showed an AICc of 2386.36. According to model 1, survival (±SE) from the ringing date (late June or July) to August was 0.81 ± 0.06, whereas monthly survival during subsequent months was 0.94 ± 0.01. The 46 estimated p values were found to range from 0.07 ± 0.06 to 0.31 ± 0.07.

Using Cormack–Jolly–Seber models that accounted for variations in the resighting probability (p) among the birds, we obtained no evidence of an effect of ring type (i.e. colour combination used on the ring) on either survival or p. Even though colour rings with black characters are often thought to be easier to read at distance than those with clear characters and a dark background, our results suggest that this bias has a negligible effect on parameter estimates from CJS models. The heterogeneity in p was probably sufficiently high to obscure any other factors that also affect this parameter, possibly including the type of ring used. For instance, some birdwatchers could be better at reading certain types of rings, while other birdwatchers could be better at reading other types of rings, so when all birdwatchers are considered together, no ring type is significantly favored in terms of readability. Moreover, some gulls are seen more frequently than others, as they visit areas close to birdwatching points (Juez et al. 2015). This causes a heterogeneity in p which is probably (as suggested in this work) of a greater magnitude than a possible effect of the type of ring on p. In this scenario, the use of different colour rings in studies, especially those relating to large-scale monitoring programs where several types of colour-ring combinations can be used, should not lead to any statistical problems. If possible, however, we recommend testing for the effect of ring type on p by adding this alternative model to the original one when utilizing CJS models.

Acknowledgments Ringing activities were authorized by the Gipuzkoa and Bizkaia regional council. This research was partly funded by the Basque Government and the Gipuzkoa and Bizkaia regional council. We are grateful to the people who helped us during the ringing and provided us with sighting data of colour-ringed gulls. The comments provided by M. Schaub and two anonymous reviewers helped us to improve an earlier version of this work.

Aradis A, Miller MW, Landucci G, Ruda P, Taddei S, Spina F (2008) Winter survival of Eurasian woodcock Scolopax rusticola in central Italy. Wildl Biol 14:36–43
Arizaga J, Galarza A, Herrero A, Hidalgo J, Aldalur A (2009) Distribucio ´n y taman ˜o de la poblacio ´n de la Gaviota Patiamarilla Larus michahellis lusitanius en el Paı ´s Vasco: tres de ´cadas de estudio. Rev Catalana d’Ornitol 25:32–42 Bermejo A, Mourin ˜o J (2003) Gaviota Patiamarilla, Larus cachinnans. In: Martı ´ R, Del Moral JC (eds) Atlas de las aves reproductoras de Espan ˜a. DGCN-SEO/BirdLife, Madrid, pp 272–273 Burnham KP, Anderson DR (1998) Model selection and inference. A practical information theoretic approach. Springer, New York Cantos FJ (2000) El anillamiento cientı ´fico en colonias de la ´ridos. Rev Anill 6:12–23 Choquet R, Lebreton J-D, Gimenez O, Reboulet A-M, Pradel R (2009) U-CARE: utilities for performing goodness of fit tests and manipulating CApture–REcapture data. Ecography 32:1071–1074 Galarza A, Herrero A, Domı ´nguez JM, Aldalur A, Arizaga J (2012) Movements of Mediterranean Yellow-legged Gulls Larus michahellis to the Bay of Biscay. Ringing Migr 27:26–31 Geroudet P (1984) Origine mediterrane ´enne confirme ´e pour les Goe ´lands leucophe ´es du Le ´man. Nos Oiseaux 37:240 Juez L, Aldalur A, Herrero A, Galarza A, Arizaga J (2015) Effect of age, colony of origin and year on survival of Yellow-Legged Gulls Larus michahellis in the Bay of Biscay. Ardeola 62:139–150 Lebreton JD, Burnham KP, Clobert J, Anderson DR (1992) Modelling survival and testing biological hypothesis using marked animals: a unified approach with case studies. Ecol Monogr 62:67–118 Molina BE (2009) Gaviota reidora, sombrı ´a y patiamarilla en Espan ˜a. Poblacio ´n en 2007–2009 y me ´todo de censo. SEO/BirdLife, Madrid Olsen KM, Larson H (2004) Gulls of Europe, Asia and North America. Christopher Helm, London Papadatou E, Pradel R, Schaub M, Dolch D, Geiger H, Iban ˜ez C, Kerth G, Popa-Lisseanu A, Schorcht W, Teubner J, Gimenez O (2012) Comparing survival among species with imperfect detection using multilevel analysis of mark-recapture data: a case study on bats. Ecography 35:153–161 Peach WJ, Siriwardena GM, Gregory RD (1999) Long-term changes in over-winter survival rates explain the decline of reed buntings Emberiza schoeniclus in Britain. J Appl Ecol 36:798–811 Pledger S, Pollock KH, Norris JL (2003) Open capture–recapture models with heterogeneity: I. Cormack–Jolly–Seber model. Biometrics 59:786–794 Pradel R, Hines JE, Lebreton JD, Nichols JD (1997) Capture– recapture survival models taking account of transients. Biometrics 53:60–72 Robinson RA, Balmer DE, Marchant JH (2008) Survival rates of hirundines in relation to British and African rainfall. Ringing Migr 24:1–6 Siriwardena GM, Baillie SR, Wilson JD (1998) Variation in the survival rates of some British passerines with respect to their population trends on farmland. Bird Study 45:276–292 Tavecchia G, Pradel R, Boy V, Johnson AR, Cezilly F (2001) Sexand age-related variation in survival and cost of first reproduction in greater flamingos. Ecology 82:165–174 White GC, Burnham KP (1999) Program MARK: survival estimation from populations of marked animals. Bird Study 46:120–139 Ye ´sou P (1991) The sympatric breeding of Larus fuscus, L. cachinnans and L. argentatus in western France. Ibis 

No hay comentarios:

Publicar un comentario