Scottish Marine and Freshwater Science Volume 5 Number 16:The Avoidance Rates of Collision Between Birds and Offshore Turbines

Appendix 1 Evidence review macro-response - barrier effect studies

A1.1 Egmond aan Zee

Location / habitat

Marine, 10-18km offshore

Turbine / array specification

Turbine array consists of 36 Vestas V90 3 MW turbines covering an area of 27 km ². Distances between turbines are 650 m within rows and 1000 m between rows. Turbine specifications given as hub height 70 m; rotor diameter 90 m; rotor altitude min 25 m (above mean sea level) and max rotor altitude 115 m (above mean sea level).

Case study number 1

Krijgsveld, K.L., Fijn, R.C., Japink, M., van Horssen, P.W., Heunks, C., Collier, M., Poot, M.J.M., Beuker, D. and Dirksen, S. 2011. Effect studies offshore windfarm Egmond aan Zee: Final report on fluxes, flight altitudes and behaviour of flying birds. Bureau Waardenburg Report No. 10 - 219.

Lindeboom, H.J., Kouwenhoven, H.J., Bergman, M.J.N., Bouma, S., Brasseur, S., Daan, R., Fijn, R.C., de Haan, D., Dirksen, S., van Hal, R., Hille Ris Lambers, R., ter Hofstede, R., Krijgsveld, K.L., Leopold, M. & Scheidat, M. 2011. Short-term ecological effects of an offshore windfarm in the Dutch coastal zone; a compilation. Environmental Research Letters 6, doi:10.1088/1748-9326/6/3/035101.

Methods

Krijgsveld et al. (2011) focussed on the disturbance of flight paths otherwise referred to as barrier effects. Whereas what was termed as the disturbance of locally resting and/or feeding birds were covered by another project (Leopold et al. 2011) as birds recorded on the water. Lindeboom et al. (2011) reported the impacts of the windfarm on a range of taxonomic groups but with respect to birds focussed on barrier effects, displacement effects and attraction. As the results presented in Lindeboom et al. (2011) were based on the preliminary results of Krijgsveld et al. (2011), cited as Krijgsveld et al. (2010), this paper is not considered further here.

Data collection was carried out during the post-construction period only.

Radar: Horizontal radar was used to record flight paths, with the radar located on a meteorological mast 500 m from the nearest turbine at the south western side of the windfarm). The radar was set to scan up to distances of 5.6 km from the meteorological mast (although it was calculated that gulls could be detected up to shorter distances of 4.5 km). There was no coverage from the angles of 155° to 220° relative to the mast however).

The radar signal was processed and recorded by Merlin (DeTect Inc). Flight paths of birds or groups of birds were visualised in QuantumGIS and grid cells (750 m x 750 m) were set up in order to analyse both the numbers of tracks and flight directions. In order to mitigate for reduced detection of tracks, due to the presence of turbines and decreasing detection rates with increasing distance from the radar, correction factors were applied to the numbers of tracks recorded inside the windfarm.

Visual and auditory observations: Panorama scans from the meteorological mast consisting of hourly 360° scans to record all birds flying within sight of the observation platform. This information was then used to calibrate the radar counts and provided information on species composition, density, flight altitude and flight direction. Additional information was collected at night and included moon watches, call registration by ear, and call registration by an automated bird call recording system. In addition, the opportunistic recording of flight paths of individual birds or bird groups (picked up either visually using a binoculars or a telescope) or on the radar) was carried out.

Study period

Radar: Continuous recording through the period of April/May 2007 to 31 May 2010. Flight path data was obtained for 817 days (out of a possible 918 due to factors such as high winds).

Visual observations : A total of 405 panoramic scans were carried out over 53 days (dawn to dusk) spread throughout the period of Feb 2006 to Dec 2009 and six nights (dusk to dawn) during spring and autumn migration (October 2008 to April 2009). Opportunistic observations of flight paths were carried between and during panoramic scans (n = 666 flight paths of 85 species were recorded with great cormorant (n = 82) and northern gannet (81) being the most commonly observed).

Species

Local seabirds (gull spp, northern gannet, scoter spp, and auks spp); migrating seabirds (diver spp and scoter spp) and migrating non-marine birds (thrushes and geese spp).

Conditions data collected under

Radar: all conditions.

Visual observations: recording carried out in generally dry, relatively calm conditions (all but day had one Beaufort scale of less than 5) and with a range of visibility conditions (0 - 50 km).

Results

Macro-responses (which were regarded by this study as being due to barrier effects), referred to in the report as macro-avoidance rates, were quantified by two methods ^[2] :

i. Panoramic scans were used to derive the proportion of birds within, at the edge and outside the windfarm. Using the combined values of the first two groupings, it was possible calculate the % of birds that passed through the windfarm ^[3] . The resulting values were corrected for relative surface area for within and outside the windfarm and then used to derive macro-avoidance rates ^[4] for northern gannet = 0.64 (n = 282 birds ^[5] ), sea ducks/scoters spp = 0.71 (n = 123 birds), diver spp = 0.68 ^[6] and alcid spp = 0.68 ³. Sample sizes were too small for other species/species groups for values to be derived and, hence, values have to be derived by other means;

ii. Flight path data collected by radar showed that the number of all birds that flew within the windfarm was on average 72% of the numbers outside the windfarm. This was proposed to equate to an average macro-avoidance rate of 0.28 of birds in relation to the windfarm, and when broken down according to time of year, the values ranged from an average of 0.18 (in winter) and 0.34 (in autumn) ^[7] . For gull spp and great cormorant, the average avoidance rate in winter of 0.18 was used, as the species composition was heavily dominated by those birds at that time of year (as shown by the visual observations). The overall average avoidance value of 0.28 was assumed for grebe spp, tubenoses spp, skua spp, and tern spp (in the absence of other available data or rationale). It was also shown using radar that the percentage of birds flying in the windfarm was significantly higher during the day compared to night (when data from spring was excluded) and these differences were greatest during summer and winter. Hence avoidance was argued to be higher at night.

Results of the opportunistic recording of flight paths indicated deflection rates of 89% for northern gannet and 40% for gulls spp based on sample sizes of 38 and 78 birds respectively ^[8] . These values were not considered by the authors to provide evidence for macro avoidance (Karen Krijgsveld pers. comm.) however.

There was inherent variation in flight direction as recorded by radar with higher variability recorded winter and summer (probably due to the inclusion of locally foraging birds which are less likely to have a consistent flight trajectory than birds migrating through the area) and during the day. Nevertheless, adjustment of flight paths occurred at 750 - 1,500 m from the windfarm when there was a pronounced change in flight direction. This was largely based on plots of the mean ± standard errors of flight direction in relation to distance according to season and time of day ^[9] . The reported changes at 750-1500 m appear to occur before and after midnight in the spring and at dusk during autumn. There were also changes in flight direction at distances further away from the windfarm but these are not highlighted - notably in spring, for most times of day, at distances between 4,500 and 5,250 m.

Numbers of birds were also shown to be highest at 750 - 1500 m, which was taken as evidence of flying birds building up as they were deflected away from the windfarm (also confirmed by visual observations of birds). Moreover, the number of tracks for all seasons in the grid cells circa 750 m from the windfarm was also shown to be significantly higher than the number of tracks for the grids cells containing the adjacent single row of turbines ^[10] .

Assessment of methodology

The values of macro-avoidance derived from the panoramic scans were species specific and were collected in a systematic manner. As for all visual observations, data collection was mostly restricted to days of reasonable visibility and calm conditions.

Macro-avoidance rates (barrier effects) derived using radar were based on mean values across all species and should be interpreted very carefully since there is likely to be variability in response rates between species. Hence this should be borne in mind when citing values derived for gull spp, grebe spp, tubenoses spp, skua spp, and tern spp. It is also unclear whether the actual numbers reported will consist of solely individual birds or whether flocks of birds may have been inadvertently included. Hence as for most radar studies, the avoidance rates cannot be necessarily assumed to correspond to those of individual birds. It is also worth bearing in mind, that the way these data have been collected (comparison of number of flight paths inside and outside the windfarm) could also be potentially considered to be evidence of displacement.

It is also problematic to overlay the arbitrarily selected boundary of 500 m buffer surrounding the outermost turbines used to delineate inside (micro and meso) and outside (macro) the windfarm avoidance ( section 3.5) with the grid cell system of 750 km ² used to analyse the number of tracks.

The grid cell system also does not correspond exactly to the boundaries of the windfarm and hence some cells will overlay areas inside and outside the windfarm which could be an issue for the values cited for % of tracks inside and outside the tracks.

A1.2 Horns Rev

Location / habitat

Horns Rev 1: Marine, 14 km offshore.

Horns Rev 2: Marine, 30 km offshore.

Turbine /array specification

Horns Rev 1: Turbine array consists of 80 2.0 MW Vesta turbines. Distance between turbines - north to south (560 m) and east to west (560 m). Turbine specifications given as: hub height 70 m; rotor blade length 40 m (diameter 80 m); and total height 110 m. Height of the lowest tip of rotor blade.

Horns Rev 2: Turbine array consists of 91 turbines. Distance between turbines - north to south (560 m) and east to west (560 m). Turbine specifications given as: hub height 68 m; rotor diameter 93 m; and total height 114.5 m. Height of the lowest tip of rotor blade 21.5 m.

Case study number 1

Petersen, I.K., Christensen, T.K., Kahlert, J., Desholm, M. & Fox, A.D. 2006. Final results of bird studies at the offshore windfarms at Nysted and Horns Rev, Denmark. Report commissioned by DONG Energy and Vattenfall A/S. National Environmental Research Institute.

Methods

This report focussed on barrier effects, displacement effects, physical changes to the habitat and collision risk. Work was carried out at the Horns Rev 1 and Nysted offshore windfarms but there were differences in methodology and timing of data collection in relation to the development phase - data collection was carried out during the post-construction period only at Horns Rev 1.

Radar observations : Recordings by radar occurred in a circular area of radius ca. 11 km (no coverage in the north east quadrant with the exception of late November 2005). The radar was located on a transformer station located less than 0.6 km from the windfarm. Migration mapped by tracing course of flocks onto a transparency and subsequently digitised. As fewer tracks were recorded both within and beyond the windfarm, due to presence of the turbines and the increasing distance from the radar, densities of tracks were not used to quantitatively to look at barrier effects.

All tracks (n = 468 north of the windfarm and n = 342 east of the windfarm) which were deemed to have a theoretical chance of entering the windfarm were selected using the criteria that they were orientated towards the windfarm at distances between 1.5 and 2 km from the windfarm and had lengths of tracks greater than 2 km.

In order to look at the lateral (horizontal) change in migration route in response (where avoidance occurs) to the windfarm, two sets of transects lines were set up. The first were located east of the windfarm running parallel to the direction of the rows of turbines (from north to south) and were set up at intervals of 50, 100, 150, 200, 250, 300, 400, 500, 1000, 2000, 2500, 3000, 3500 and 4000 m (max. range set by limits of the radar). The second were set up north of the windfarm at 50, 100, 200, 300, 400, 500, 1000, 1500, 2000, 2,500, 3000 m and then at intervals of 1000 m until 7000 m. The orientation of all bird tracks that intersected two adjacent transects were calculated for all of the transects running east and north of the windfarm.

Visual observations: four transects from the transformer station set up, one of which passed diagonally through the windfarm.

Study period

Radar observations: A total of 17 survey periods (shortest = 5 h 30 min, longest = 39 h 30 min) were carried out covering the periods of August to November 2003; March to May 2004; August to September 2004; March to May 2005; and August to November 2005. Total of 243 h 45 min of observations.

Visual observations: 19 surveys (shortest = 7 h 0 min, longest = 29 h 30 min) were carried out covering the periods of April to May 2003; August- November 2003; March to May2004; August to September 2004; March to May 2005; and August to November 2005. Total of 403 h 18 min of observations.

Species

Staging and migrating birds. Based on visual observations of birds during transect counts, likely to consist primarily of diving ducks (by an order of magnitude higher than any other group and consisting almost exclusively of common scoter), gulls (herring gull, little gull, greater back-backed gull and black-legged kittiwake and terns (Sandwich tern and common/Arctic tern) ^[11] .

Conditions data collected under

During day and night, weather conditions not presented.

Results

The annual percentage of bird tracks (based on the years 2003, 2004 and 2005 ) entering the windfarms from either the northern or the eastern side of the windfarm ranged from 13.6 % (2005, north of windfarm) and 29.3% (2004, east of the windfarm ^[12] ). The number of tracks that these percentages are based upon are relatively small however (ranging from 12 to 39 tracks). These values appear to provide the origins of cited macro-avoidance rates of 0.71 and 0.86. Spring and autumn periods were not differentiated between as it was argued that bird behaviour would be similar regardless of the time of season.

The mean orientation of tracks of migrating birds, as calculated for all intervals between transects, was used as the response variable to look at the lateral deflection of south bound tracks for birds north (n = 2108) and east of the windfarm (n = 1168). For birds north of the windfarm during southbound bird migration, analyses ^[13] showed that distance to windfarm, wind direction (crosswinds), time of day and the interaction between distance and time of day were significant. Plots of the mean flight orientation with distance to windfarm in relation to time of day wind direction showed that deflections were most pronounced at distances of less than 400 m from the windfarm and that changes at larger distances (<2 km) were more obvious during the daytime compared to the night time period ^[14] . For birds east of the windfarm analyses ^[15] found that distance had a significant effect on the orientation of the birds (wind direction, time of day and the interaction between distance to windfarm and wind direction were also significant. Plots of the mean flight orientation with distance to windfarm in relation to time of day wind direction showed that deflections were most pronounced at distances of less than 500 m from the windfarm. Changes in orientation occurred up to 4 km from the windfarm during south bound migrations notably during the day in westerly winds ^[16] .

Assessment of methodology

The derived macro-avoidance rates (based on barrier effects) are a mean value for all birds which occurred during the study and according to visual observations consisted mainly of common scoter. Therefore, these reported avoidance rates may have limited applicability to the less commonly recorded gulls spp and tern spp. In addition these avoidance rates are based on relatively small sample of tracks. Moreover, tracks do not differentiate between individuals or flocks, therefore the reported macro-avoidance rates do not respond to the level of individual birds.

Case study number 2

Blew, J., Hoffman, M., Nehls, G. & Hennig, V. 2008. Investigations of the bird collision risk and the responses of harbour porpoises in the offshore windfarms Horns Rev, North Sea, and Nysted, Baltic Sea, in Denmark. Part 1: Birds. Report from the University of Hamburg and BioConsult SH, 145pp.

Methods

The report focussed on the collision risk to migrating birds at Horns Rev 1 and Nysted offshore windfarms and the same methodology was used at both sites.

Blew et al. (2008) proposed that avoidance occurred at the three broad scales of: (1) large scale avoidance >2000 m; (2) medium to small scale avoidance 1000 m to 150 m and either horizontally or vertically as measured directly (reactions) or indirectly (comparison of numbers or flight altitudes); (3) last second avoidance. Thus, the second category, which was the focus of this report, overlaps with the definitions in section 3 of this report of both macro- and meso-avoidance.

Data collection was carried out during the post-construction period only.

Radar observations : Horizontal radar (Bridgemaster E-series and Pathfinder) was deployed from ships with a range of anchoring sites (three, four and four at the eastern, southern and western edges of the windfarm respectively) at distances of 150 to 300 m to the windfarm. Screenshots were captured using a digital camera for the horizontal radar and the angle of tracks and their length were also registered. The range of the radar was set to 1.5 nautical miles. No manual tracking of signals on the horizontal radar was carried out which meant that changes in flight trajectories for individual tracks could not be looked at.

Radar tracks were categorised according to their direction in relation to the first row of the windfarm; flying towards (± 45° either side of perpendicular to the windfarm; flying away; and flying parallel (more or less).

In order to look at lateral avoidance, four intervals ranging from 0-500 m, 500-1,000 m, 1,000-1,500 m and 1,500-2,000 m in relation to the ship and the relative orientation of tracks were recorded in the range of ± 90° with 0° being perpendicular to the windfarm. Due to sample size issues (insufficient number of tracks), it was not possible to report results for Horns Rev, however.

Visual observations: Visual observations were carried out along a 2 km transect which ran perpendicular to the outer edge of the windfarm, with the ship located halfway along it length. On the windfarm side of the transect, the gap between the edge of windfarm as defined by the row of the outer turbines (approximately 300 m from the ship) to 700 m inside the windfarm (or 1,000 m from the ship) was regarded as being inside the windfarm. On the corresponding non-windfarm side, the transect which was between 300-1,000 m from the ship was regarded as being outside the windfarm (in relation to the windfarm this represents a distance of between 600 and 1,300 m). Collectively these were termed as Class A, whereas the transect up to 300 m either side of the ship was Class B (excluding birds within 30 m either side of the ship which were disregarded). Visual observations of flying birds (optics only used for identification purposes) were carried out every half hour for observation periods of 15 minutes from sunrise to sunset. Distance, flight direction and altitude were recorded (classes were largely defined by the upper and lower limits of the rotor blade: 0-5 m: 5-30 m; 30-100 m; >110 m). The results of this work are not considered further here.

Visual observations were carried out for 219.5 and 238.5 h in 2005 and 2006 respectively.

Study period

March to May to coincide with spring migration (27.5 observation days in 2005 and 2006) and September to November to cover autumn migration (39 observation days in 2005 and 2006).

Radar appeared to have been run continuously.

Species

Seaducks, geese, gulls and terns and wide range of songbird species.

Transect counts showed that gulls (many of which were unidentified to the species level) were the most common group recorded in both spring and autumn (with little gull notably more common in the former time of year). Common scoter were also common but more so in spring.

Conditions data collected under

Horizontal radar observations were limited to calm sea state conditions (wind speed < 2 ms ^-1) and generally dry weather.

Visual observations were stopped when visibility <1 km but visual and acoustic observations were possible for all observation days

Results

During the day, the overall number of tracks flying parallel to the windfarm was higher (n = 1,045) compared to flying away from (n = 486) or towards (n = 386) the windfarm. This pattern was less pronounced at night with the number of birds parallel to the windfarm (n = 253) being only marginally higher compared to flying away from the windfarm (n = 206) but were higher than towards the windfarm (n = 101).

Although the visual observations were designed primarily to look at the differences in flight height distribution, they were able to provide supporting evidence for macro avoidance occurring. For northern gannet, out of 66 gannets recorded only 2 flew within the windfarm. For both little gull and all gull spp (excluding little gull), significantly less birds were present inside the windfarm.

Assessment of methodology

Results from the observations from horizontal radar were limited as only 5% (9% for Nysted) of the observation time yielded screenshots which could be used and these were biased to daytime periods. There was also the additional problem that detection within the windfarm was considerably lower compared to outside due to the presence of the wind turbines (tracks were observed to disappear and reappear when entering and leaving the windfarm).

There were several limitations with working on a ship compared to from land or a fixed platform, including rough sea conditions, which would likely hamper data collection. There were also issues associated with the tidal cycles (particularly at Horns Rev, less so at Nysted) and strong winds which could result in the ship turning and this affected the radar data collected. Another potentially confounding factor is that the ship could also act as an attractant to some species of seabirds ( e.g. gull spp) or potentially act as a disturbance to others ( e.g. diver spp and duck spp).

In terms of demonstrating macro-avoidance, horizontal radar was unable to provide quantitative evidence. Avoidance appeared to be implied by the percentage of birds flying parallel being higher than those values reported for birds flying towards and away from the windfarm and this pattern was more pronounced during the day when the windfarm was more visible. The significance of birds tracks running parallel to as opposed to being orientated towards or away from the windfarm was not explained, however, and there was a lack of pre-construction information to make comparisons with. There was also insufficient data to look at potential changes in the orientation of tracks (but enough data was available for Nysted - see section 5.4.4). Similarly the visual observations did not provide quantitative evidence of macro-avoidance rates.

Case study number 3

Skov H., Leonhard, S.B., Heinänen, S., Zydelis, R., Jensen, N.E., Durinck, J., Johansen, T.W., Jensen, B.P., Hansen, B.L., Piper, W. & Grøn, P.N. 2012. Horns Rev 2 Monitoring 2010-2012. Migrating Birds. Orbicon, DHI, Marine Observers and Biola. Report commissioned by DONG Energy.

Methods

This report focussed on migrating birds in relation to Horns Rev 1 and 2.

Radar observations: Horizontal radar was used from observation stations located to the north east of Horns Rev 1 (assumed to be the same as used in previous studies at Horns Rev 1, 560 m distance to the windfarm) and to the east of Horns Rev 2 (no distance provided but estimated to be less than 2 km away). Radar range was set at 6.0 km and covered a circular area. Additional information on species identification was possible by use of "a real-time tracking" procedure whereby tracks of individual birds or tracks could be followed on background images to produce videos. Videos were produced using a frame grabber connected to the radar and tailor made software provided the video as a back ground image on the PC screen. Whilst one observer followed the trace on the screen, a second attempted to locate the target in the field using a binocular or telescope to provide names, number of birds and altitude. Identification on tracks was not always possible during busy periods. Track densities were estimated for a 100 m ² grid system within the radius of the radar.

Laser range finders: Laser range finders (Vectronix 21 Aero) were also used from the observation stations used to collect species-specific data up to distances of 2-3 km for large bird species (depending on the field of view and flight mode of the bird). Positions and altitudes of birds were logged automatically via GPS recorded at intervals of 10-15 sec. Data from the laser range finders were used to supplement data collected by the radar. Calibrations in order to correct the readings provided by the GPS were necessary due to interference by the observation tower.

Track data for range finders and radar were also integrated with weather data including wind direction, wind speed, air pressure, clearness, humidity, total precipitation and air temperature. In addition, the relative flight direction of the bird in relation to wind direction was also calculated.

Generalised Additive Models ( GAMs) with a Tweedie distribution were used to look at track densities derived by radar for all bird tracks and common scoter tracks in relation to distance to the radar and distance to the windfarm. Generalized Additive Mixed Models ( GAMMs) with a correlation structure (to deal with spatial and temporal autocorrelation) were used to look at the flight altitude in relation to weather variables and distance to the nearest wind turbine. However, this information could not be used to quantify an avoidance rate.

Study period

Data collection carried out during spring and autumn from September 2010 to May 2012. No further details given.

Radar observations: 15 min per h during daylight.

Laser range finders: operated permanently with observation periods of a minimum of 15 min per h.

Species

All spring and autumn migrants (seabirds, water birds, ducks and passerines).

Conditions data collected under

Not specified.

Results

Tracks recorded by both horizontal radar and the laser range finders were mapped for a range of species/groups in order to visualise movement patterns. It was proposed that diver spp (small sample size), northern gannet and common scoter tended to migrate along corridors along the periphery of the windfarms, although looking at the maps provided it is clear that northern gannet ^[17] and common scoter ^[18] did occur within the windfarms, notably Horns Rev 2. This was thought to be a result of the bathymetry as common scoters seemed to associate with waters less than 10 m in depth.

At Horns Rev 2 both distance to radar and distance to the windfarm were significant predictors of the densities for all birds tracks combined ^[19] and common scoter tracks. Response curves ^[20] produced by the models were similar for both analyses, which was unsurprising given the relative proportion of all tracks that were from common scoter. A peak in the density of birds occurred at around 1,500-2,500 m from the windfarm and was argued to provide evidence for a barrier effect due to birds altering their flight path. Similarly at Horns Rev 1, both distance to radar and distance to the windfarm were significant predictors for all bird tracks and common scoter tracks. In terms of the response curves, distance to windfarm the peak for all birds was between 2,000-3,000 m, whereas for common scoter it was around 1,000-2,000 m ^[21] .

Assessment of methodology

From the results provided it is not possible to quantify an overall macro-avoidance rate although this study did provide information on the distances to which barrier effects were observed.

A1.3 Nysted offshore Windfarm

Location / habitat

Marine, offshore 10 km.

Turbine /array specification

Turbine array consists of 72 2.3 MW Bonus turbines covering 24 km ². Distance between turbines - north to south (480 m) and east to west (850 m). Turbine specifications given as: hub height 69 m; rotor blade length 41 m; total height 110 m. Clearance above water is 28 m.

Case study number 1

Petersen, I.K., Christensen, T.K., Kahlert, J., Desholm, M. & Fox, A.D. 2006. Final results of bird studies at the offshore windfarms at Nysted and Horns Rev, Denmark. Commissioned by DONG Energy and Vattenfall A/S. National Environmental Research Institute.

Desholm, M. & Kahlert, J. 2005. Avian collision risk at an offshore windfarm. Biology Letters 1: 296-298 ^[22] .

Methods

Peterson et al. (2006) focussed on barrier effects, displacement effects, physical changes to the habitat and collision risk. Work was carried out at Horns Rev and Nysted offshore windfarm but there were differences in methodology and timing of data collection. Study at Nysted covered the three phases of: baseline (1999-2002); during construction (2002-2003) and post-construction (2003-2005). Desholm and Kahlert (2005) reported the results from the barrier effects and collision risk work only.

Radar observations: Recordings by radar (Furuno FR125) were carried out from an observation tower, 5 km north-east of the windfarm area. The range was approximately 11 km and covered a circular area of 388 km ². Migration was mapped by tracing the course of flocks onto a transparency and subsequently digitised. Only tracks longer than 5 km were included in the analyses.

The lateral response to the windfarms was investigated by setting a number of transects: the eastern gate (located along the full length the most eastern edge of the windfarm); the northern gate (located along the full length the most northern edge of the windfarm) and the buoy transect (running from north to south from the observation tower to a buoy, 6.9 km in length). During autumn migration, tracks of flocks of birds travelling in a westerly direction which crossed the buoy transect were selected to see if they crossed the eastern gate (in order to derive the percentage of birds which did so). In contrast, during spring migration the flight behaviour of birds was studied after they passed the windfarm and so is not considered further here. The total numbers of flocks of birds crossing the eastern and northern gate were also counted. In addition, migration intensities were compared for an area within the windfarm with an adjacent area outwith the windfarm (both less than 11 km ² in area). Each area was subdivided into squares of 0.1 km ² and within each cell, the lengths of radar tracks (bird flocks) were expressed as the total sum of track meters (the track density). In order to derive the change due to the windfarm, proportional differences in the bird densities within and inside the windfarm from the baseline data (pre-construction) were used to correct the data collected post-construction to derive avoidance rates.

In order to determine the response distance (where avoidance occurs) to the windfarm, transect lines to the east of the windfarm were set up which ran parallel to the direction of the rows of turbines (from north to south). These were spaced at intervals of 100, 200, 300, 400, 500 m and then at intervals of 500 m to 4,000 m and after which there were a further two transects at 5,000 and 6,000 m. The mean ± s.d. migration course of tracks were calculated for each transect (based on the gap between the transect itself and the 100 m interval to the west).

Visual observations : Abundance, phenology, diurnal pattern and flock sizes of species were recorded along the buoy transect. Count data was then converted into number of birds per 15 mins for all westerly bound birds in autumn and easterly bound birds in spring (although again the latter represents the number of birds after passing through the windfarm).

Study period

Radar observations: spring (easterly-orientated migration) and autumn (westerly-orientated migration) periods covered. Total number of hours or breakdown by season not reported.

Visual observations : During the main survey periods of 14 March to 19 April and 30 August to 12 November from 1999 - 2005, observations were carried out two days per week covering day and night time periods. A total of 259 h and 579 h observations gathered for the spring and autumn periods.

Species

Staging and migrating birds but common eider and geese spp most commonly recorded.

Conditions data collected under

Not specifically described but very little data of conditions under poor visibility (<1 km).

Results

The probability of birds crossing the windfarm was analysed using a logistic regression model and included the following explanatory terms and first order interactions (phase of development; distance to the observation tower when crossing the buoy transect), time of day, direction of winds (all of which were found to have significant effects). It was shown through comparison of data from the baseline and operation phases that 0.78 of all birds ^[23] , which consisted mostly of common eider, avoided entering the windfarm post-construction during autumn migration. This was based on 40% of flocks entering the eastern edge of the windfarm during the baseline period compared to 9% during operation ^[24] . This was suggested to equate to 8 out of 10 flocks crossing the eastern gate during the baseline study then avoiding the windfarm during the post-construction phase. It was also shown that during the post-construction phase, the numbers of flocks crossing the eastern gate were higher at night than during the day (Desholm and Kahlert 2005 cited values of 13.5 % and 4.5 % respectively).

More specifically there was notable inter-annual variation in macro-avoidance rates for autumn migrating birds, again mostly common eider, ranging from 0.63 and 0.83 ^[25] in the use of the windfarm post-construction compared to the baseline. These rates were derived from figures of 0.08-0.09 of flocks passing the eastern side of the windfarm compared to 0.24-0.48 passing the eastern gate of the windfarm during the pre-construction period ^[26] .

There was a difference in migration intensity during the baseline period as the track densities in the eastern windfarm were 60% of the reference area which suggested a problem with detection rate. Nevertheless a significant reduction in track densities was reported for the post-construction period but there was acknowledgement that a reduction could be partially explained by problems of what is termed a shadow effect to do with individual turbines.

The standard deviation of the orientation was used to determine the lateral deflection as means of quantifying response distance to the windfarm (citing Kahlert et al. (2005) as justification for this approach). Analyses of data collected during the autumn migration, showed a significant interaction between the phase of development and distance to the windfarm (other terms were also significant but not discussed here due to lack of information presented which can be evaluated with respect to providing evidence for the response distance) ^[27] . Plots of the means of annual standard deviation values showed that there was little change in orientation for distances between 100 m and 5 km from the windfarm during the baseline period ^[28] . However, during the operation period, the orientation of tracks steadily changed over the distances 5 to 1 km away from the windfarm (orientation of birds at 3 km from the windfarm were significantly different to the baseline period) and the greatest deflection occurred between 500 m and 100 m (note that the way the transects were set up, there was a gap between 500 m and 1 km). A tendency was also reported for the first deflection to be recorded at greater distances during the day compared to the night time period (based on the multiple use of pair-wise t tests across each distance interval) ^[29] .

Assessment of methodology

As there was a before and after comparison carried out at Nysted this was argued to provide greater confidence (compared to Horns Rev) that any changes were as a direct result of the windfarm presence.

The response distance was only possible for birds entering the windfarm during autumn (the area used during spring migration was beyond the edge of the radar range and hence the derived figures are based on autumn migration only. Moreover, tracks do not differentiate between individuals or flocks, therefore the reported macro-avoidance values do not respond to the level of individual birds.

Case study number 2

Blew, J., Hoffman, M., Nehls, G. & Hennig, V. 2008. Investigations of the bird collision risk and the responses of harbour porpoises in the offshore windfarms Horns Rev, North Sea, and Nysted, Baltic Sea, in Denmark. Part 1: Birds. Report from the University of Hamburg and BioConsult SH, 145pp.

Methods

Methods used were exactly the same as used for Horns Rev ( Appendix 1, section A1.2)

Study period

March to May to coincide with spring migration (44 ship days in 2005 and 2006) and September to November to cover autumn migration (51.5 ship days in 2005 and 2006).

Radar appeared to have been run continuously.

Species

Wide range of non-pelagic waterbirds with high numbers of common eider as well as raptors and songbirds. Transect counts showed that in spring, the common eider was by far the most common bird recorded and in autumn it was the great cormorant.

Conditions data collected under

Horizontal radar observations were limited to calm sea state conditions (wind speed < 2 ms ^-1) and generally dry weather. Weather and sea state conditions tended to be better than those experienced at Horns Rev where fewer observation days were possible.

Visual observations were stopped when visibility <1 km.

Results

Radar tracks were categorised according to their direction in relation to the first row of the windfarm: flying towards (± 45° either side of perpendicular to the windfarm; flying away; and flying parallel (more or less). Initially tracks were presented regardless of their location (and therefore distance) in relation to the windfarm (but included tracks within the boundary of the outer row of the windfarm). During the day the overall number of tracks flying parallel to the windfarm was higher (n = 2,274) compared to towards (n = 1,725) or away (n = 563) from the windfarm. This pattern was not evident at night when the numbers flying towards (n = 968) and parallel (n = 804) were more similar but still much higher than flying away (n = 216).

In terms of determining whether horizontal avoidance occurred, the mean (and standard deviations) of angles of the approaching tracks were presented for the four 500 m width distance bands, for all anchor points east and west of Nysted offshore windfarm. It was reported that the angles did not increase (as would be predicted if horizontal avoidance occurred) or differ with decreasing distance to the windfarm (no statistical analyses were carried out).

Although the visual observations were designed primarily to look at the differences in flight height distribution, they were able to provide supporting evidence for macro avoidance occurring. For all gull spp significantly less birds were present inside the windfarm. No results for northern gannet were provided.

Assessment of methodology

See Appendix 1, section A1.2 for a discussion regarding the work carried out on radar and visual observations at Horns Rev where the same approach was used. With respect to looking for evidence of horizontal avoidance this study was unable to show evidence for a change in flight orientation. It was unclear though whether this was due to relatively wide bands being used (500 m in width) as other studies have used smaller intervals of 100 m at distances less than 1,000 m from the windfarm.