Abstract
This report contains the results of the extended baseline monitoring on bird migration in relation to the planned Horns Rev 2 offshore wind farm (HR2 OWF) undertaken September-November 2008. The extended baseline on bird migration followed up on baseline monitoring activities undertaken during spring 2008 to boost the cumulative database on bird migration across Horns Rev which was judged insufficient for the baseline of the HR2 monitoring programme. The insufficient data available on long-distance migration of waterbirds across Horns Rev is largely due to the methodological constraints and limitations using visual observations and radar screen analyses for quantifying bird migration and the inherently low samples obtainable from ship-based radar installations. Additionally, the monitoring programme for Horns Rev 1 (HR1) OWF did not cover the central and western areas of Horns Rev. Thus, an extended monitoring programme was established during autumn 2008 using fixed radar installations at Horns Rev 2 (foundation), the Horns Rev 1 Transformer Station and at Blåvands Huk. Visual observations were carried out from the HR1 transformer station and from Blåvands Huk, and were intended to provide calibration data for classification of the radar data into bird species groups. A total of 824 bird observations were made at HR1 and 2,554 at Blåvands Huk.
The three installations were based on the LAWR system design, which use software developed by DHI for high-resolution signal processing, data extraction, automatic classification and GIS-interfacing. The LAWR is based on X-band technology, using standard marine radars, type FR2127 from Furuno. The data acquisition hardware allows sampling of up to 24 images per minute, which facilitates object tracking. All radar equipment includes ancillary hardware linked to the systems, allowing 24 hour operation and remote control. The radar software is subdivided into 3 parts: RadCtrl2/PolScan (radar control and acquisition), BirdWatch/BirdWatchShow (on-line ground truth data collection) and BirdTrack (classification and extraction of bird tracks). Wind information and the presence of rainfall in the radar coverage area have been estimated and included in the radar image database. The echoes were sampled at 20 MHz at 10 bit resolution (1,023 levels) and collected in “bins” each covering a radial distance of 120 m and 1 degree tangentially. The sample time for one image was one minute. For further processing the mean, peak and variance of the radar signals were calculated “on-the-fly” at the data collection computer at each radar station. Volume- and en-route correction of the echo was handled using the standard correction scheme used on the LAWR during the last 10 years. Identification of tracks was made using algorithms operating on 120 successive radar images. From the recorded wind-speed and winddirection and the corresponding data for the track bird heading and velocity (speed through air) was calculated. Filtering procedures were established which efficiently distinguishes echoes from potential birds from other echoes like ships, rain and wind-induced clutter.
A post-classification routine was developed to estimate the statistical probability for a trajectory belonging to a certain class of birds, which was made by applying a classification tree model, which was established based on the calibration data set. The classification tree compared classification potentials of bird classes separated by expected air speed, size and flight characteristics of the track. The selected classification tree model was deployed on the combined radar tracks resulting in the additions of estimated bird classes to each track. They were subsequently used to generate spatial trends and time series profiles of flight intensities (frequency of tracks) for each bird class.
The sample size of potential bird flight trajectories was 535,482 of which 145,918 were from the radar at Blåvands Huk, 278,078 from HR1 and 111,486 from HR2. The classification tree analysis generated a ‘simple’ tree with only two splits resulting in three terminal nodes dominated by large waterbirds, ducks and passerines. The variance of the echo reflectance was the most important factor separating the three bird classes, followed by the speed through air. The modelled flight intensities showed similar trends across the three bird classes. The flight intensities of both waterbirds and landbirds strongly indicated higher passage rates of birds in the eastern-most sector of Horns Rev, significantly lower rates at the HR1 OWF and rather low rates at HR2; the area to the east of the planned HR2 wind farm had slightly higher rates than the area to the west of the planned wind farm. These results point at the existence of a marked migration corridor in autumn of approximately 10 km width at the inner Horns Rev – Blåvands Huk area.
Although no information on altitudes were collected during this study the information collected on flight intensities reveal potentials for avoidance behaviour of sea birds and collision risks associated with the construction of the HR2 OWF. Given the results of the trends and profiles of migration intensities of the three major bird groups there are no indications of large-scale migration occurring at HR2 in autumn. Additionally, no short-term events were noted at the site. The trends in flight intensities at HR1 OWF indicated avoidance response by all three bird classes: ‘large waterbirds’, ‘ducks’ and ‘passerines’. The spatial patterns of flight intensities recorded during this investigation show that the reduced flight intensities are completely related to the wind farm and the surrounding area to a distance of approximately 1.5 km, and there are no indications of e.g. shading effects on the southern fringe of the wind farm. Thus, a local avoidance effect at HR1 OWF must be expected to impact the southbound migration of a wide range of species, albeit the effect in terms of modified flight paths and energetic costs are deemed to be minor for long-distance migrants. As the results show that HR1 OWF is at the western margin of the migration corridor in autumn the effects is most likely to cause long-distance migrants moving towards the Wadden Sea to adjust eastwards and migrants with a south and southwesterly course to temporarily split paths when passing the wind farm. Due to the short and local scale of avoidance effect it is not likely that the moderate displacement of migrants at HR1 will affect the migrants en-route to HR2 OWF, and the effect of avoidance at HR2 OWF is therefore most likely to be local (non-cumulative). The recorded migration intensity of the two bird classes ‘large waterbirds’ and ‘ducks’ was slightly higher on the eastern than on the western side of HR2 OWF, while the intensities of ‘passerines’ were equal on both sides. Thus, the avoidance effect at HR2 is likely to display both a concentration of birds (non-passerines) on the eastern side and a split of migrating passerines around the wind farm.
The application of automated recordings from fixed radar installations at Horns Rev 2, the Horns Rev 1 Transformer Station and Blåvands Huk in combination with visual calibration observations has resulted in a vastly increased database on bird migration along the Horns Rev area. The collected data are judged sufficient to establish a baseline for future monitoring of bird migration during the autumn season.