Fully accounting for nest age reduces bias when quantifying nest survival

Abstract

Accurately measuring nest survival is challenging because nests must be discovered to be monitored, but nests are typically not found on the first day of the nesting interval. Studies of nest survival therefore often monitor a sample that overrepresents older nests. To account for this sampling bias, a daily survival rate (DSR) is estimated and then used to calculate nest survival to the end of the interval. However, estimates of DSR (and thus nest survival) can still be biased if DSR changes with nest age and nests are not found at age 0. Including nest age as a covariate of DSR and carefully considering the method of estimating nest survival can prevent such biases, but many published studies have not fully accounted for changes in DSR with nest age. I used a simulation study to quantify biases in estimates of nest survival resulting from changes in DSR with nest age under a variety of scenarios. I tested four methods of estimating nest survival from the simulated datasets and evaluated the bias and variance of each estimate. Nest survival estimates were often strongly biased when DSR varied with age but DSR was assumed to be constant, as well as when the model included age as a covariate but calculated nest survival from DSR at the mean monitored nest age (the method typically used in previous studies). In contrast, biases were usually avoided when nest survival was calculated as the product of age-specific estimates of DSR across the full nesting interval. However, the unbiased estimates often showed large variance, especially when few nests were found at young ages. Future field studies can maximize the accuracy and precision of nest survival estimates by aiming to find nests at young ages, including age as a covariate in the DSR model, and calculating nest survival as the product of age-specific estimates of DSR when DSR changes with nest age. LAY SUMMARY: Nest survival rates are useful tools for predicting population trends and understanding habitat quality for birds. However, nest survival rates can be difficult to measure accurately in the wild. Nest survival rates can be biased if nests are not discovered by researchers on the first day of the nesting interval (age 0) or if changes in daily survival probability during the nesting interval are not fully accounted for when describing cumulative nest survival from hatching to fledging. This study simulated nest data from known (true) survival rates to compare four methods of estimating nest survival to hatching or fledging. Each estimate was compared to the true value to measure the accuracy of each method. In the simulation, when all nests were found after age 0 and the daily survival rate changed with age, estimates of nest survival to hatching or fledging were often biased. Biases were prevented by (1) estimating how daily nest survival rates changed with age, rather than ignoring age; and (2) calculating the probability of a nest surviving to hatching or fledging as the product of age-specific daily survival rates. This study demonstrates the importance of fully accounting for nest age and provides a framework for future studies to do so.