|
Résumé :
|
Recent evidence indicates that molt-dependent sampling bias may distort estimates of population-level molt phenology, but these distortions can be minimized by making explicit use of data derived from individuals recaptured at different points of the same molt cycle. We investigated these issues by first examining the pattern of molt-dependent sampling bias in 26 passerine species mist-netted during active flight-feather molt at Powdermill Avian Research Center in southwestern Pennsylvania, USA. We then created simulated datasets of birds in active molt that modeled different types of sampling bias and examined the ability of two conventional and two recapture-based methods to produce unbiased and accurate estimates of molt phenology: Pimm regression (PR), the Underhill-Zucchini type 3 model (UZ3), a recently developed recapture-based extension of UZ3 (recapUZ3), and a novel mixed-model extension of Pimm regression that incorporates data from recaptured individuals (recapPR). Molt-dependent sampling bias was widespread in the Powdermill data and usually took the form of bimodal sampling bias, whereby birds in the early and late stages of flight-feather molt were disproportionately sampled, probably because impaired flight performance during the intense middle stages of molt decreases the probability of capture by mist net. When sampling bias is molt-dependent, recapPR and PR produce unbiased estimates of molt start and duration, and recapPR provides increased accuracy as the number of recaptured individuals increases; estimates derived from Underhill-Zucchini methods, in contrast, are either biased (UZ3) or inaccurate when the number of recaptured individuals is small (recapUZ3). However, when sampling bias is date-dependent—that is, when sampling effort is not uniformly distributed throughout the period of active molt—recapPR and PR produce biased or inaccurate estimates, and recapUZ3 or UZ3 are better analytical options. Our findings emphasize the importance of molt- and date-dependent sampling bias and the value of recapture data in estimating molt phenology and clarify the conditions under which conventional and recapture-based versions of Pimm regression and Underhill-Zucchini type 3 methods perform best and are recommended.
|
