Comparing pre- and post-construction distributions of long-tailed ducks around an offshore wind farm: A quasi-designed experiment accounting for imperfect detection, local surface features and autocorrelation

We describe a technique to model abundance patterns of wintering seaducks in relation to the construction of an offshore wind farm (OWF) based on seven years of aerial survey transect data. Distance sampling was used to estimate seaduck densities adjusted for covariates affecting detection probabilities. A generalized additive model (GAM) generated seaduck densities in sampling units in relation to spatially explicit covariates, using bootstrapping to account for uncertainties in both processes. Generalized estimating equations generated precision measures for the GAM robust to spatial and temporal autocorrelation. Comparison of pre- and post-construction model generated surfaces showed significant reductions in long-tailed duck numbers only within the OWF (despite the fact that the model was uninformed about the OWF location), although the absolute numbers involved were small at the population level. This method provides quantification of distributional effects on organisms across the survey area and time as an alternative to Before-After/Control-Impact designs in environmental impact assessment