Data from: Vegetation phenology and nest survival: diagnosing heterogeneous effects through time

Birds should select nest sites that minimize predation risk, but understanding the influence of vegetation on nest survival has proven problematic. Specifically, the common practice of measuring vegetation on nest fate date can overestimate its effect on survival, simply because vegetation at hatched nests grows for a longer period of time than vegetation at nests that were depredated. Here, we sampled the literature to determine the prevalence of this bias in studies of duck breeding ecology. We then used survival data collected from ~2,800 duck nests to empirically evaluate evidence of bias in four different vegetation metrics: vegetation density measured when the nest was found, density when the nest was fated, and date-corrected regression residuals of these two. We also diagnosed the magnitude of vegetation effects on nest survival by restricting analysis to only nests which were fated contemporaneously (thereby removing potential bias in the timing of measurement). Finally, we examined whether systematic phenological differences exist between vegetation at hatched and depredated nests that have the potential to further obfuscate the relationship between vegetation and nest survival. We found evidence for a true positive effect of vegetation density on nest survival that appeared to be inflated when using raw vegetation measurements collected at fate date. However, taken in combination with the literature review, our results suggest that the majority of duck nesting studies have evaluated the role of vegetation on nest survival using a relatively less biased metric—vegetation density when the nest was found. Finally, we found that over the course of a nesting attempt, vegetation increased in density at successful nests, but decreased in density at depredated nests. This seasonal change potentially points to an important new metric for understanding predation risk, but further experimental research is required to fully eliminate potential biases in the timing of vegetation measurements.