Abstract
The main aim of the EOWDC Bird Collision Avoidance Study has been to improve our understanding of seabird flight behaviour inside an offshore wind farm. This should be achieved through collection of as detailed seabird flight data as possible rather than through estimation of avoidance rates for collision risk modelling per se. The focus is on seabird flight behaviour during the breeding period and post-breeding period when densities are highest in the Aberdeen area. The technical improvements of the monitoring equipment employed in the Aberdeen Offshore Wind Farm made it possible to track seabirds inside the array and measure meso-avoidance more confidently than before. It has been possible to match video camera recordings of seabird movements to a sample of their radar tracks. A total of 1,753 coupled tracks were recorded during 2020 and 1,370 tracks during 2021, which was beyond expectations and formed the basis for robust assessments of flight behaviours of target species in different parts of the wind farm array. The target sample size for species-specific meso-avoidance of 250 was reached for all key species, and the target for micro-avoidance of 100 was reached for herring gull (Larus argentatus) and black-legged kittiwake (Rissa tridactyla, hereafter referred to as kittiwake). The level of mesoavoidance recorded was 0.5 for kittiwakes, 0.7 for herring gulls and 0.5 for Northern gannets (Morus bassanus, hereafter referred to as gannet) and great black-backed gulls (Larus marinus). Together with the recorded high levels of micro-avoidance in all target species (> 0.96) it is now evident that seabirds will be exposed to very low risks of collision in offshore wind farms during daylight hours. This was also substantiated by the fact that no collisions or even narrow escapes were recorded in over 10,000 bird videos during the two years of monitoring covering the April – October period.
Detailed statistical analyses of the seabird flight data were enabled both by the large sample sizes and by the high temporal resolution in the combined radar track and video camera data (2.5 seconds). The flight data were analysed in relation to the local wind and turbulence (wake) conditions using multivariate Random Forest models. The most accurate video data were collected during 2021 when the video tracker on both cameras was upgraded to a new version which uses deep learning algorithms to separate flying seabirds from other flying objects and has the capability to keep the tracked bird in the centre of the field of view and record the tracks for longer periods. The target species displayed horizontal meso-avoidance within 100-120 m distance from the rotors. Herring gulls showed maximum meso-avoidance of 0.7 close to the blades, and kittiwakes showed avoidance of 0.5. As expected, they also displayed attraction to the areas in between the turbine rows. Gannets and great black-backed gulls only displayed avoidance at distances closer than 40 m and 50 m, respectively from the tip of the rotor blades. Both species displayed an avoidance level of 0.5.
For comparison with the ORJIP BCA study meso avoidance rates were estimated in the EOWDC study using the same algorithm and the results showed comparable rates with ORJIP for unidentified large gulls and gannet, while slightly lower rates were estimated for herring gulls and great black-backed gulls, and much lower rates were calculated for kittiwakes.
The results from the EOWDC study strongly indicate that the within wind farm avoidance response of the studied species of seabirds towards turbines mainly takes place within 100-120 m distance from rotors and that the response intensifies as the seabirds approach the rotor blades. In proximity to the rotors the recorded meso-avoidance response behaviour for all four species was manifested as a complex 3-dimensional pattern. Commuting gannets appeared to reduce flight altitude, whereas herring gulls and kittiwakes displayed a slight increase in mean flight height as they approached the rotor blades. When assessing the recorded flight orientation of the birds relative to the rotors commuting gannets and kittiwakes appeared to deflect around 80 m distance from the rotors and herring gulls at 50 m. The flight models revealed that turbulence and wind speed had the strongest effect on the profiles of flight behaviour of all target species. The pattern of responsive flight behaviour seemed to break down during situations with strong turbulence, while wind speed mainly affected the distance at which the increase in flight height took place. The trends resolved by the flight models were apparent irrespective of whether the birds were recorded as feeding or commuting.
The tendency to deflect and fly parallel to the rotor means that although meso-avoidance seems to be lower than anticipated prior to this study, micro-avoidance is very strong as seabirds are rarely recorded crossing the spinning rotors without adjustments as captured by the analyses of microavoidance. Large gulls (herring gull, lesser black-backed gull (Larus fuscus) and great blackbacked gull) adjusted their flight behaviour to cross the rotor either obliquely or perpendicularly more frequently than gannets and kittiwakes. The recorded micro-avoidance rates (> 0.96) are similar to the micro-avoidance rate (0.957 ± 0.115 SD) which was estimated for large gulls in the ORJIP project using the same methods as in this project (Skov et al. 2018). These flight characteristics translate into very low risk of collision.
Despite the evidence of the low risk of collision by seabirds in the EOWDC, the Random Forest flight models revealed that the mean avoidance response pattern may break down during specific weather conditions. The model results indicate that all four target species show different flight profiles towards the rotor.