Abstract
This document is a memo that is written as an addition to the KEC 4.0 assessment reports of the effects of collision risk (Potiek et al. 2022) and habitat loss (Soudijn et al. 2022) from offshore windfarms on seabird populations. It relies heavily on the methodology described in these reports and cannot be considered independently from these reports. Because of uncertainties in the density maps of the herring gull as used in the KEC 4.0 assessment, exploratory analyses were conducted to improve the maps. New national density maps were created based on five methods and were compared with the original herring gull density maps from KEC 4.0. The density maps based on predictions of a regression model with the covariates distance to coast, fishing intensity and water depth as predictors seem to represent the natural (mainly coastal) distribution of the herring gull the best. Regression models are capable of handling data with many zero observations (“zero-inflated” data) and high variability in observations of especially high bird densities, which is typical for the herring gulls at sea. Based on the original maps and the maps newly developed here, the estimated population-level effects of collision mortality exceed the ALIs for all national scenarios. Yet, the estimated collision mortality was found to be slightly lower with the new density maps based on the regression model. Due to the strong time constraints of the project, the newly developed density maps need to be considered as preliminary. Future research should focus on model selection to tease out the most appropriate covariates to predict the herring gull distribution and developing a better proxy for fishing activity. In addition, the confidence intervals estimated in the densities by the regression model should be used to indicate uncertainties in the outcomes of the casualty estimations and population-level effects. Finally, rigorous testing of the performance of the model (through a comparison of model predictions with field observations) should be conducted in order to determine its predictive power.