Abstract
A substantial part of the nocturnal bird migration in Europe passes through the North and the Baltic Sea. These migrating birds are increasingly confronted with development of offshore wind farms, which pose a risk of collision with these anthropogenic structures. It is commonly assumed that the number of collisions increases with the number of birds flying at rotor height. This assumption would mean that collision fatalities peak during periods of peak migration traffic rates (MTR). However, due to lack of suitable methods at sea, empirical data are lacking.
The main aim of the present study was to quantify the collision risk of migrating birds and to test the hypothesis of a strong positive relationship between collision rate and MTR. Alternatively, a bird’s collision risk may depend on other factors, foremost weather conditions. For example, adverse weather may reduce the ability of a bird to perceive the rotor in time to avoid it.
With a combination of innovative and established methods, we recorded the MTR of birds at a coastal onshore wind farm using a specialised bird radar system (BirdScan MR1 by Swiss Birdradar Solution AG) and recorded the number of rotor transits, defined as birds crossing the rotor plane, using AI-supported camera systems (AVES Offshore HPC by ProTecBird GmbH) during more than three complete migration periods. This enabled us to analyse the relationship between the number of collisions and MTR. Using these data, we determined the avoidance rates of nocturnal and diurnal migrants - a parameter of paramount importance in collision risk models (CRM) that captures the proportion of birds avoiding the rotors when approaching the wind farm. Applying stochastic Band CRMs, along the exact MTR at rotor height and empirically estimated avoidance rates, we then calculated the expected number of collisions for the whole study period (including both day and night activity).
Finally, to validate the theoretically calculated number of collision fatalities, we employed an independent and well-established empirical method. Using thorough Post Construction Fatality Monitoring (PCFM), we searched for bird carcasses at the turbines at a 5-day interval and estimated the total number of fatalities while correcting for search area, searcher efficiency and carcass persistence.
In general, the number of rotor transits was very low when the turbines were operational. During night, on average one transit was recorded every 132 hours when the rotation speed was ≥2 rounds per minute. However, there was a remarkable difference in rotor transit rates depending on the operational status of the rotors. When turbines were inactive, transit rates were about 20 times higher than when rotors were operating, suggesting a reduced avoidance response at idle turbines. Furthermore, our results did not show a strong positive relationship between rotor transit rates and MTR. In other words, even during nights of high migration intensities, the probability of collisions to occur did not increase. However, contrary to our expectation, weather parameters also explained only a small part of the probability of rotor transits and therefore collision risk. These results indicate that collision risk may depend mostly on other factors, which remain to be investigated.
Avoidance rates, calculated by comparing the number of observed rotor transits with the expected number based on MTR at rotor height, were found to be high: 0.9987 during the night and 0.9986 during the day when the turbines were operating. This means, an estimated 99.87 and 99.86 percent of birds, approaching the wind farm at rotor height while turbines were operating, avoided the rotor plane during night and day, respectively. Also, the overall estimated collision risk for birds flying through the wind farm area at 25–1025 m altitude was very low: for nocturnal migrants only 0.0016 percent of flights were expected to result in a collision, for daytime movements the risk was similarly low at 0.0020 percent. Comparison of the theoretically calculated number of collisions with the empirically determined fatalities using PCFM, showed a high overall agreement for the period of this study (Band CRM: 76.6 fatalities [95%CI: 57.3-97.6] and PCFM: 99.7 [95%CI: 55.0-168.0]). In addition, collision victims found during the main bird migration seasons did not contain species known to constitute the bulk of nocturnal migration at the study site. This i) confirms the overall low collision risk for migrating birds, ii) validates estimated rotor transits and avoidance rates and iii) highlights the usefulness of applying CRMs to estimate collision risk when fed with realistic avoidance rates and appropriate site-specific turbine and bird flux rate data.
These results have important implications for potential measures to mitigate the risk of collision for migrating birds, such as turbine curtailment during periods of high migration intensity. Such curtailment measures imply a strong positive relationship between the probability of bird collisions and migration intensity, which could not be confirmed by our findings. Therefore, turbine curtailment during periods of high migration intensity is likely to be ineffective.