Handling of uncertain nest fates and variation in nest survival estimates

Abstract

Researchers are moving their attention away from the Mayfield method of estimating nest success to advanced nest survival models that can be implemented within statistical software packages. In spite of this, little attention has been paid to developing formal rules for handling of ambiguous nesting data within these software-based methods. I compared the variation caused by differential handling of data between the hand-calculated Mayfield estimates and those obtained using the constant survival logistic-exposure method. Four variants of each of the two methods were applied to sets of nest records (n = 5,476) of nine open-nesting passerines. Of all nest fates, 57% (unweighted mean across species) were categorized as failed, 29% as successful and 14% as uncertain, according to either age criterion or combination of all fate evidence criteria. Different methods yielded survival estimates that differed as much as 6% over a 25-day nesting cycle. Variation in logistic-exposure survival estimates caused by the four variants of interval coding was higher (range = 4.4%) than variation in the Mayfield estimates (range = 2.7%) caused by the four variants of exposure termination. Researchers are urged to consider different handling of ambiguous nesting data as one of the many possible sources of bias when implementing any method of nest survival analysis.