Abstracts

Presented by Christian Roy - Email: christian.roy3@canada.ca

We evaluated double-dependent observer methods for helicopter surveys as a mean to adjust counts of waterfowl for incomplete detection. We tested our methodology during a sea duck survey in Labrador, eastern Canada, in 2009 and subsequently applied it to a larger sea duck monitoring project in the Northwest Territories and Manitoba, Canada, in 2017 and 2018. We used a hierarchical Bayesian model with a data augmentation scheme that allowed us to derive detection rates for birds, true bird abundance, group size, and group composition (i.e. male, female, and unknown). We estimated detection rates by crews, and within crews by observer position (front vs. rear) and experience. We used the output of the model to derive the predicted number of total indicated breeding pairs for the survey area. Detection rate was highest from the backseat of the helicopter despite a reduced field of vision and experienced observers had a slightly better detection rate. There were individual differences in detection with experienced observers tending to have higher detection probabilities than less experienced ones. Detection probabilities also varied among species with higher detection rates for sea ducks and divers and lower for dabblers and geese. Overall detection rate was generally high, but failing to correct counts for incomplete detection could lead to an underestimation of total indicated pairs by 5% for some species. We recommend that doubleobserver approaches be used in future helicopter surveys to reduce bias associated with observer turnover and allow a measure of detection probabilities.