Estimating nests not present at the time of breeding surveys: An important consideration in assessing nesting populations

Counts of nest starts are often used as indicators of the size of avian nesting populations, or of avian productivity. However, the accuracy of single or repeated counts of unmarked nests over time for estimating seasonal numbers of nests may be strongly affected by nest events that fall in between survey dates, or that occurred prior to or after the survey period. Accuracy may also be affected by uncertainty in the interpretation of counts due to overlap between starting and ending dates of asynchronous nests during the intervisit interval. To measure the combined magnitude of these effects on survey accuracy, we overlaid a monthly "survey" regime on known initiation and ending dates of 2055 nests of ciconiiform birds. Assuming all nests present on the date of simulated survey were counted, monthly surveys underestimated the true number of nest starts by 24-64%, depending on species and year. Using a simple model, we also demonstrate that accuracy does not increase much as survey frequency increases, and that significant estimation error can occur over a wide range of nest success values and degrees of asynchrony. We suggest that (1) these biases can be significant for surveys of many kinds of nesting birds including some territorial passerines, (2) this bias cannot be addressed by increasing survey frequency, and (3) the degree of renesting may be of critical interest for inferring breeding population size from nest count data. We suggest three possible approaches for modeling this error.