Collision risk of Montagu's Harriers Circus pygargus with wind turbines derived from high‐
resolution GPS tracking
Schaub, Tonio; Klaassen, Raymond; Bouten, Willem; Schlaich, Almut E.; Koks, Ben J.
Published in:IBIS – International Journal of Ornithology DOI:
10.1111/ibi.12788
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date: 2020
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
Schaub, T., Klaassen, R., Bouten, W., Schlaich, A. E., & Koks, B. J. (2020). Collision risk of Montagu's Harriers Circus pygargus with wind turbines derived from high‐resolution GPS tracking. IBIS – International Journal of Ornithology, 162(2), 520-534. https://doi.org/10.1111/ibi.12788
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.
Collision risk of Montagu’s Harriers Circus pygargus
with wind turbines derived from high-resolution GPS
tracking
TONIO SCHAUB1,2* RAYMOND H.G. KLAASSEN,1,3WILLEM BOUTEN,4ALMUT E. SCHLAICH1,3& BEN J. KOKS1
1
Dutch Montagu's Harrier Foundation, Postbus 46, 9679 ZG, Scheemda, The Netherlands
2
Animal Ecology, Institute of Biochemistry and Biology, University of Potsdam, Maulbeerallee 1, 14469, Potsdam, Germany
3
Conservation Ecology Group, Groningen Institute for Evolutionary Life Sciences, University of Groningen, Postbus 11103, 9700 CC, Groningen, The Netherlands
4
Computational Geo-Ecology, Institute for Biodiversity and Ecosystem Dynamics, University of Amsterdam, Postbus 94248, 1090 GE, Amsterdam, The Netherlands
Flight behaviour characteristics such asflight altitude and avoidance behaviour determine the species-specific collision risk of birds with wind turbines. However, traditional obser-vational methods exhibit limited positional accuracy. High-resolution GPS telemetry rep-resents a promising method to overcome this drawback. In this study, we used three-dimensional GPS tracking data including high-accuracy tracks recorded at 3-s intervals to investigate the collision risk of breeding male Montagu’s Harriers Circus pygargus in the Dutch–German border region. Avoidance of wind turbines was quantified by a novel approach comparing observed flights to a null model of random flight behaviour. On average, Montagu’s Harriers spent as much as 8.2 h per day in flight. Most flights were at low altitude, with only 7.1% within the average rotor height range (RHR; 45–125 m). Montagu’s Harriers showed significant avoidance behaviour, approaching turbines less often than expected, particularly whenflying within the RHR (avoidance rate of 93.5%). For the present state, with wind farms situated on the fringes of the regional nesting range, collision risk models based on our new insights onflight behaviour indicated 0.6– 2.0 yearly collisions of adult males (as compared with a population size of c. 40 pairs). However, the erection of a new wind farm inside the core breeding area could markedly increase mortality (up to 9.7 yearly collisions). If repowering of the wind farms was car-ried out using low-reaching modern turbines (RHR 36–150 m), mortality would more than double, whereas it would stay approximately constant if higher turbines (RHR 86– 200 m) were used. Our study demonstrates the great potential of high-resolution GPS tracking for collision risk assessments. The resulting information on collision-relatedflight behaviour allows for performing detailed scenario analyses on wind farm siting and tur-bine design, in contrast to current environmental assessment practices. With regard to Montagu’s Harriers, we conclude that although the deployment of higher wind turbines represents an opportunity to reduce collision risk for this species, precluding wind energy developments in core breeding areas remains the most important mitigation measure. Keywords: avoidance rate, environmental impact, flight height, human–wildlife conflict, mitigation, raptors, renewable energy, wind energy.
Worldwide, wind-power generation has grown rapidly during the last two decades, as it represents a key component of the efforts to limit climate change (Tabassum-Abbasi et al. 2014). However,
*Corresponding author.
Email: tonio.schaub@grauwekiekendief.nl Twitter: @grauwekiek
concerns about negative impacts of wind energy infrastructure on populations of wild animals have been raised, including mortality through collisions with the rotor blades of wind turbines in birds and bats (Schuster et al. 2015). Among birds, raptors are known to be especially collision-prone (Barrios & Rodrıguez 2004), and serious impacts on regio-nal raptor populations have been found in several cases (Dahl et al. 2012, Bellebaum et al. 2013). Yet, at present, knowledge of the risk of collision with wind turbines and associated population effects is incomplete for many species. It is there-fore highly desirable to improve knowledge on particular aspects offlight behaviour, such as flight altitude and avoidance behaviour, which deter-mine species-specific collision risk.
To date, collision-related flight characteristics have mostly been assessed based on visual observa-tions (Dahlet al. 2013, Hull & Muir 2013), which are subject to limited three-dimensional positional accuracy and weather bias (Johnston et al. 2014). GPS telemetry represents a promising method to overcome this, as individual birds can be followed over extended periods regardless of weather condi-tions (Ross-Smith et al. 2016, Peron et al. 2017). However, the location error in GPS measurements can be considerable, which is problematic when flight altitude and distance between birds and wind turbines are to be assessed. On the one hand, the error may be tackled by applying sophisticated modelling techniques (Ross-Smith et al. 2016, Peron et al. 2017). On the other hand, the posi-tional precision may be strongly improved by increasing the GPS fix frequency (Bouten et al. 2013, Corman & Garthe 2014, Thaxter et al. 2018), which offers unprecedented opportunities forfine-scale, three-dimensional analyses. Here, we used such ‘high-resolution’ GPS tracking data, recorded at afix interval of 3 s, to explore the col-lision risk of an endangered raptor species.
Among the different aspects of flight behaviour related to the collision risk with wind turbines, the birds’ avoidance behaviour on multiple scales (Cook et al. 2014) is particularly important but, at the same time, difficult to study (Chamberlain et al. 2006). Typically, avoidance behaviour has been quantified by comparing the expected num-ber of collisions in the absence of avoidance beha-viour derived from collision risk models (CRMs; e.g. Band et al. 2007) with the actual number of collision fatalities recorded with carcass searches (Vasilakiset al. 2016). However, instead of relying
on the bias-prone results from carcass searches (Madders & Whitfield 2006), we based our analy-sis of avoidance behaviour exclusively on flights recorded by GPS tracking in this study, comparing observed tracks with a null model of random flight behaviour (Hull & Muir 2013, Thaxter et al. 2018).
We studied a breeding population of Montagu’s Harrier Circus pygargus, a medium-sized, migra-tory bird of prey, in the northern Dutch–German border area. Montagu’s Harrier is Red-listed in both countries because, due to intensive agricul-ture, its populations depend today largely on con-servation management (active protection of nests in agricultural crops and agri-environmental schemes for enhancing prey availability; Trier-weiler 2010, Schlaich et al. 2015). Wind energy represents a potential additional threat, as Mon-tagu’s Harriers occur in open landscapes which are attractive for wind energy developments. Indeed, occasional collision fatalities of this species have been reported throughout Europe (D€urr 2017).
First, we aimed to improve our understanding of the flight behaviour of Montagu’s Harriers by describing general flight activity (proportion of time spent in flight), the extent to which wind farm areas were visited (‘horizontal overlap’), the frequency distribution of flight altitude (‘vertical overlap’ with wind turbine rotors) and the birds’ avoidance of turbines. Secondly, as a case study, we estimated the mortality of Montagu’s Harriers caused by wind turbines in the study area using ‘Band’ CRMs (Band et al. 2007) based on the above-mentioned flight characteristics, in combina-tion with the regional distribucombina-tion of nest-sites. Thirdly, we compared mortality through collisions in different scenarios of wind energy expansion in the study area, specifically repowering and the construction of a new wind farm, to identify measures to reduce collision risk for Montagu’s Harriers.
METHODS
Study area and wind farms
Our study area was situated in the north-east of the Netherlands and in adjacent areas of north-western Germany. The relatively uniform, flat and open landscape was built up of polders – land reclaimed from the sea, surrounded by dikes – where land use was dominated by intensive
agriculture, mainly cereal production. The area supported a well-defined breeding population of 30–50 pairs of Montagu’s Harriers.
Several wind farms were situated on the fringes of the study area, three of which were regularly visited by Montagu’s Harriers (Delfzijl Zuid, Bunde, Rhede; Fig. 1; Table S1 in Appendix S1). To define ‘wind farm areas’, we drew a concave poly-gon through the outer wind turbine towers of the wind farms and added a buffer of 200 m (Vasilakis et al. 2016; Fig. 1). The Bunde wind farm was repowered with larger turbines in 2016 (Table S1 in Appendix S1).
Technical characteristics of the wind turbines were obtained from the wind farm operators or from publicly available documents. For all analyses except stage II of the ‘Band’ CRM (see below), a buffer of 5 m was applied around the rotor in the vertical plane allowing for imprecision in the recorded flight altitude (see below) and for the fact that smaller birds such as Montagu’s Harriers may die without being directly hit (‘barotrauma’; O. Krone pers. comm.). The ‘rotor height range’ (RHR) was defined to be the range between the minimum and the maximum rotor tip height (low-est and high(low-est point the rotor tips can take), including the buffer of 5 m in both directions. For general inspection of the vertical overlap of Mon-tagu’s Harriers’ flights with wind turbines, we applied the ‘average rotor height range’ of the wind turbine models present in the study area (45–125 m above ground level).
GPS tracking
In the context of a long-term conservation pro-gramme, 33 Montagu’s Harriers were equipped with solar-powered UvA-BiTS GPS trackers (Bou-tenet al. 2013; www.uva-bits.nl) in the study area between 2009 and 2016 (Schlaich et al. 2015). Birds were trapped close to their nest either with a mist-net in combination with a stuffed raptor or with a snare-trap mounted on a perch. GPS track-ers were mounted using a full-body harness made from 6-mm-wide Teflon ribbon strings (Kenward 1987). Trackers, including harness, weighed approximately 12 g (on average 4.7% of the birds’ body weight; range: 3.6–6.1%). Data were remo-tely downloaded from the trackers using the UvA-BiTS antenna system (Bouten et al. 2013). There was no indication of reduced breeding perfor-mance in tagged birds (Table S2 in Appendix S1).
Due to sample size limitations in females (seven female individuals were tracked, but the breeding behaviour of females hampered battery charging so that only small amounts of data could be col-lected), the study was restricted to male breeding birds. The overall dataset used in this study com-prised data from 24 Montagu’s Harrier males (two non-breeding males excluded), recorded on 2559 individual tracking days (Table 1). See Appendix S1b for details on data selection.
The trackers collected three-dimensional GPS positions (longitude, latitude, altitude above sea level) every 5 min (‘low-resolution data’). Addition-ally, ‘high-resolution’ tracking data with a GPS fix interval of 3 s were collected, mostly during periods of 1 h per day which were regularly shifted across daytime hours to prevent bias (see Appendix S1b and c for details). To a lesser extent, ‘virtual geo-graphical fences’ (defined areas where GPS trackers automatically change settings) were used to collect high-resolution tracking data inside particular areas of interest such as agricultural fields with specific agri-environmental schemes (Schlaich et al. 2015) or wind farms (< 10% of high-resolution data col-lected based on fences). In total, the high-resolution dataset included 944.0 h of flight tracks from 22 birds (Table 1).
The mean error of UvA-BiTS GPS trackers was found to amount to < 3 m in terms of both hori-zontal position and altitude in high-resolution data, compared with 20–70 m in low-resolution data (Bouten et al. 2013). Additionally, we assessed the vertical precision of our own tracking data based on the recorded altitude above ground level for periods when the birds were stationary (see below), when the true altitude was expected to be close to zero (see Appendix S3 for details). The mean vertical error was 3.34 m in the high-resolu-tion data (95% quantile: 8 m) and 13.40 m in the low-resolution data (95% quantile: 35 m).
For every GPS position, the trackers also recorded instantaneous speed (2D and 3D) and positional dilution of precision (PDOP), which were used to classify and filter the data (see below). The PDOP indicates the positional accu-racy of GPS fixes based on the geometry of avail-able satellites (Langley 1999).
Data analysis
All data processing and statistical analyses were performed in R version 3.4.2 (R Core Team
2017). For all spatial analyses and to create maps, we used the R packages sp (Pebesma & Bivand 2005), OpenStreetMap (Fellows 2016) and OSMs-cale (Boessenkool 2017).
Flight activity
We distinguished between stationary and in-flight GPS fixes based on the two-dimensional
instantaneous speed. This variable was bimodally distributed and, assuming that the two peaks rep-resent stationary moments on the one hand and in-flight moments on the other, we applied the between-peak minimum as a threshold separately for every individual (individual thresholds ranging from 1.1 to 2.9 m/s for low-resolution data and from 0.5 to 1.5 m/s for high-resolution data).
Figure 1. Map of the study area indicating present and planned wind farms (state of 2014), the distribution of nest-sites of Montagu's Harriers in 2009–2016 (a; n = 279 nests) and the tracks and nest-sites of the GPS-tagged birds (b; n = 40 nests from 24 birds). Small map, bottom left: situation of the study area within the Netherlands. Figures on the right: spatial configuration of the three wind farms which were most often visited by the tracked birds. The tracking data in (b) were sub-sampled to a 15-min interval. Tiles of background maps by Stamen Design, under CC BY 3.0. Map data by OpenStreetMap, under ODbL. Coordinates of centre of maps (a) and (b): 53.12°N, 7.08°E. [Colour figure can be viewed at wileyonlinelibrary.com]
Table 1. Overview of the datasets used for different parts of the analysis.
Aspect offlight
behaviour Data selection
No. of GPS
fixes Time span No. of
birds
No. of bird-seasons
Flight activity Low resolution 346 116 2559 days 24 40
Horizontal overlap
Low resolution and high resolution sub-sampled to 5 min, inflight 173 675 2554 days 24 40 Vertical
overlap
High resolution, inflight 875 322 944.0 h 22 34
Avoidance behaviour
High resolution, coherentflight tracks of at least 10 GPS fixes inside WF, within a minimum distance of turbinesa
8057 8.7 h 5 7
The proportion of time the birds spent in flight was assessed using low-resolution data (Table 1) because of a more equal seasonal and diurnal spread of data compared with high-resolution data. We divided the day into hourly bins, calculated the number of GPS fixes in flight divided by the total number offixes within each bin (all individu-als combined throughout the breeding season) and summed these hourly proportions to obtain an overall average of time spent in flight per day in hours.
Horizontal overlap with wind farms
We calculated the proportion of GPSfixes in flight that lay inside wind farm areas per ‘bird-season’ (n = 40), i.e. for every individual during every breeding season. For this analysis, we used both low- and high-resolution data, the latter subsam-pled to the measurement interval of low-resolution data (5 min; Table 1).
In some cases, birds breeding in or near wind farms were specifically selected for tagging, which implied that the tracked birds were not entirely representative of the population with regard to the distance between their nest-sites and wind farms (7.5% of nest-sites of tagged birds were within 2 km of the Bunde wind farm compared with 4.7% of all nest-sites, 7.5% compared with 5.4% for Delfzijl Zuid). To correct for this bias, we built regression models based on the tracked birds, which we used to predict the average proportion of time spent in the wind farms for the popula-tion. Binomial generalized linear models (GLMs) were constructed separately for each of the three regularly visited wind farms with the ratio between the number of GPS fixes inside and the number of fixes outside the respective wind farm per bird-season as response and nest-to-wind-farm distance as predictor variable using the R function glm. The other wind farms (Fig. 1) were visited to a negligible extent (see below) and were therefore not considered. We chose this simplistic modelling approach without random effects (year or individ-ual) because inclusion of random effects resulted in largely unrealistic regression curves, probably due to the small and imbalanced sample of indi-viduals (relatively few birds visiting the wind farms, small variation in nest-to-wind-farm dis-tances within individuals). We used Bayesian methods to derive 95% credible intervals for the model coefficients (Korner-Nievergelt et al. 2015). We applied the function sim from the R package
arm (Gelman & Su 2015) to obtain a sample of 5000 simulated values from the joint posterior dis-tribution of the model coefficients. The mean of the 5000 simulated values was used as estimate; the 2.5% and 97.5% quantiles were used as lower and upper limits of the credible interval. A coeffi-cient was considered significantly different from zero if its credible interval did not include zero. Note that due to the simplified model structure (see above), there is an increased risk of Type I error in the models.
We used information about nest locations on the population scale (279 nest locations from years 2009–2016; Dutch Montagu’s Harrier Foundation pers. comm.; Fig. 1a) in conjunction with the above-mentioned regression models to predict the average proportion of time that a male Montagu’s Harrier spends in the wind farms for use in the collision risk models (see below). The range of nest-to-wind-farm distances from all nests only slightly exceeded the range of distances from the nests of the tracked birds that were used to build the model (0.1–33.0 km compared with 0.1– 24.6 km).
Vertical overlap with wind turbine rotors
To analyse the flight altitude of Montagu’s Harri-ers, we exclusively relied on high-resolution track-ing data (Table 1) to maximize positional precision. Altitude above ground level was deter-mined by subtracting the elevation of the terrain above sea level, which was retrieved from the SRTM3 global elevation model (resolution of 90 9 90 m; Jarvis et al. 2008), from the bird’s altitude above sea level provided by the GPSfixes. Due to the very flat topography in the study area (95% between 5 and 0 m above sea level), we expect only low inaccuracy in flight altitudes above ground level resulting from the resolution of SRTM3.
Even in the high-resolution data, negative alti-tudes above ground level were present to a consid-erable extent (21.4% of fixes), 93.2% of which fell between 10 and 0 m. This was probably due to the generally low flight altitude of Harriers, which can easily lead to negative altitudes even with low imprecision in the GPS altitude measurements and low inaccuracy of the ground elevation model. These negative flight altitudes were not excluded per se in order not to introduce a skewed bias. Instead, we omitted particularly imprecise fixes based on the positional dilution of precision
(fixes with PDOP > 4 excluded; D’Eon & Delparte 2005).
To assess the effect of wind turbine dimensions on vertical overlap, we considered different rotor radii (20–70 m with increments of 10 m) in com-bination with a fine scale of minimum rotor tip height (15–145 m with increments of 1 m), cover-ing the range of conceivable turbine models. (The range of rotor radii of turbine models in the study area was 20–57 m, and the range of minimum rotor tip heights was 17–97 m.) For every combi-nation, we determined the proportion of GPSfixes within the RHR (all GPS fixes used regardless of whether they were located inside or outside wind farms).
Avoidance behaviour
For the analysis of avoidance behaviour, the data-set was restricted to high-resolution tracks of at least 10 consecutive GPS fixes inside wind farms, resulting in a sample of 217 tracks from five indi-vidual Montagu’s Harriers, consisting of 8057 GPS fixes (Table 1; see Fig. S4 in Appendix S2 for maps). Different ways of obtaining a null model of random space use that incorporates movement constraints are conceivable (e.g. rotate and/or shift original tracks, construct correlated random walks; Richard et al. 2012, Calenge 2015). Here, we opted for horizontally rotating the original tracks with a random angle around their centroids in order to maintain their internal structure and their approximate location within the wind farm. We built‘sets’ of simulated tracks, each containing one rotated track per observed track (Fig. 2). We cre-ated 1000 of these simulcre-ated sets in total (see Appendix S1e for details).
To derive avoidance rates per altitude class (be-low, within and above the RHR), we determined the proportion offixes within a horizontal distance of less than the rotor radius from the nearest tur-bine tower (‘risk distance’) in the original data and in every simulated set. The avoidance rate (AR) was calculated as AR¼propexppropobs
propexp (with propexp = mean of the 1000 expected proportions; propobs = observed proportion). The AR ranged
between∞ and 1, with positive values indicating avoidance and negative values indicating attraction. Statistical significance for the hypothesis of the observed value being smaller than the expected values was derived using a permutation test (one-sided, significance level of 5%) implemented by
the function as.randtest in the R package ade4 (Dray & Dufour 2007). The AR within the RHR was used in the CRMs (see below).
We investigated how far avoidance behaviour extended away from the turbines by dividing the range of distances to the nearest turbine into bins of 20 m. We determined the proportion of fixes within every distance bin both in each simulated set and in the original data per altitude band, sepa-rately for the two individuals with the largest sam-ple sizes (representing 82.3% of the dataset; the other three birds did not have sufficient sample sizes in all altitude bands). We computed an avoidance/attraction index (AAI) per bin by scal-ing the difference between propobs and propexp
with the average between the observed and the mean expected proportion (propobs=exp; see Fig. S2 in Appendix S1 for an illustration): AAI¼propobspropexp
propobs=exp . Negative AAI values indi-cate avoidance, positive values attraction. Of the 1000 AAI values per bin, we derived the mean and the 95% range (2.5–97.5% quantiles). The 95% range represents the acceptance range of the null hypothesis of the observed value being differ-ent from the expected values (two-sided test); if it does not include zero, avoidance/attraction is sig-nificant at the 5% level.
Collision risk models
For the three most visited wind farms, we calcu-lated the expected number of turbine collisions of Montagu’s Harrier males in the present state using ‘Band’ CRMs (Band et al. 2007). Note that for the Bunde wind farm, we applied the state of the wind farm before repowering for this analysis. Stage I of the ‘Band’ model estimates the number of rotor crossings assuming absence of avoidance beha-viour; stage II calculates the probability of colliding for a single rotor crossing. The results of these two stages are multiplied by each other and by the rate of non-avoidance (1 – avoidance rate) to obtain the expected number of collisions. Similar to the approach of Vasilakis et al. (2016), we multiplied the average number of breeding pairs of Montagu’s Harriers in the study area (indicating the popula-tion size of breeding males) by the length of the breeding season in days, the amount of time spent in flight per day, the average proportion of time spent in the respective wind farm (as predicted from the regression models on horizontal overlap) and the proportion of time spent within the RHR
to obtain the overall amount of time the male population spends inside the ‘wind farm risk vol-ume’ (RVwf; Table S3 in Appendix S1) per year.
Regarding the proportion of time spent within the RHR, we applied the overall average across all tracking data because flights inside and outside wind farms did not differ in flight altitude (Appendix S4). The rest of the calculation fol-lowed the guidelines in Band et al. (2007). The time spent inside the combined volume swept out by the rotors (‘rotor risk volume’, RVrot; Table S3
in Appendix S1) was calculated by multiplying the time spent in RVwf by the ratio RVrot/RVwf.
Sub-sequently, the number of transits through the rotor was calculated by dividing the time spent in RVrot by the time needed to cross RVrot once
(see Appendix S1f for details). For stage II, we applied a publicly available Microsoft Excel spreadsheet (Band 2017).
On the one hand, we applied the avoidance rate determined in the present study (see above). On the other, we used the default value of 98% proposed in case of missing empirical evidence (Scottish Natural Heritage 2016). See Appendix S1g for an overview of the input param-eters used in the CRMs and the respective sources. The number of annual collisions predicted by the CRMs was divided by the population size of Mon-tagu’s Harrier males to obtain the ‘additional annual per-capita mortality’.
Besides the calculation for the present state, we applied three scenarios of wind energy expansion in the study area: (1) ‘Repowering low’, with all turbines of the three wind farms being replaced by
the modern turbine model set in place during the actual repowering of the Bunde wind farm (3.4 MW, RHR 36–150 m; Table S1 in Appendix S1), keeping the number of turbines constant; and (2) ‘Repowering high’, using a variant of the model used in (1) with 50-m higher towers (RHR 86– 200 m), which is also commercially available. With these repowering scenarios, the total rated power capacity of the wind farms was doubled compared with the original state. (3) Thirdly, a hypothetical new wind farm of the shape of Delfzijl Zuid was projected to variable locations across the study area (scenario ‘New wind farm’). A grid with an edge length of 1 km was laid over the study area, with the grid nodes representing the centroids of the new wind farms. All parameters were adopted from the present-state CRM for Delfzijl Zuid except the average proportion of time spent inside the wind farm, which was predicted individually for every grid node from the binomial GLM based on the distances of nest-sites to the new wind farm location (see above). The mean number of nest-sites within 2 km of the centroids of the hypotheti-cal wind farms per year ranged between 0.0 and 5.9. Note that we did not consider the effect of habitat suitability in this simplified model, assum-ing that the new wind farm has the same habitat configuration as Delfzijl Zuid, which implies that the mortality estimates for the hypothetical wind farm locations must not be taken at face value. Fur-thermore, based on results of an earlier study (Hernandez-Pliego et al. 2015), we assumed that the distribution of nest-sites would be unaffected by the wind farm development.
(a) (b) (c)
Figure 2. Illustration of the rotation approach applied to create randomflight trajectories. (a) Example of original track (grey line) and corresponding simulated track (green line) obtained by rotating the original track by a random anglea around its centroid (star). (b,c) Sets of original and simulated tracks of one individual Montagu's Harrier. Open circles: wind turbine locations. Rectangular shape in right part of (b): parcel of extensive grassland which was frequently visited by this individual. [Colourfigure can be viewed at wileyon linelibrary.com]
R ES UL TS Flight activity
The GPS-tracked Montagu’s Harrier males spent on average 8.2 h per day in flight. This corre-sponds to 47.5% of daytime spentflying (applying an average day length of 17.3 h).
Horizontal overlap with wind farms In total, nine wind farms were visited by the tracked birds. Only Delfzijl Zuid, Bunde and Rhede yielded more than 1% of in-flight GPS fixes inside the wind farm for at least one bird-season. The proportion of fixes inside wind farms varied con-siderably between individuals, with an average of 2.0% (1.6% forDelfzijl Zuid, 0.3% for Bunde, 0.1% for Rhede). The maximum value per bird-season was 34.8%, stemming from an individual breeding inside the Delfzijl Zuid wind farm. For all three regularly visited wind farms, a significant negative relationship between the proportion of GPS fixes inside the respective area and the nest-to-wind-farm distance was found (Table 2), and the strength of the effect varied considerably between the wind farms (Table 2, Fig. S3 in Appendix S2). When using these models to predict the average proportion of time spent inside the three wind farms for the population of Montagu’s Harriers, an estimate of 0.96% of time spent inside was obtained for Delfzijl Zuid, compared with 0.18% and 0.05% forBunde and Rhede, respectively. Vertical overlap with wind turbine rotors Most flights of Montagu’s Harriers as recorded by high-resolution GPS tracking were close to the ground, with 7.1% of in-flight GPS fixes within the average RHR (Fig. 3). Combining these results with the daily amount of time spent in flight yielded 34.9 min spent within the average RHR per day.
The proportion of fixes within the RHR increased exponentially with decreasing minimum rotor tip height at constant rotor radius (Fig. 4). For example, the new turbines in Bunde after repowering (Table S1 in Appendix S1) yielded a 2.0-fold increase compared with the original tur-bines (Fig. 4).
Avoidance behaviour
When flying below the RHR, the tracked Mon-tagu’s Harriers approached the turbines relatively closely (shortest horizontal distance to a turbine tower: 7.4 m). Still, significant avoidance beha-viour towards turbines was present (avoidance rate of 25.6%; Table 3, Fig. 5). Within the RHR, avoid-ance behaviour was more pronounced and extended further away from the turbines (until 60–80 m; Fig. 5), the shortest recorded distance being 20.5 m. The avoidance rate amounted to 93.5%.Above rotor height, no significant avoidance behaviour was exhibited (Table 3, Fig. 5). Here, the shortest distance to a turbine was 7.0 m. Collision risk models
The CRMs for the present state with the default avoidance rate of 98% predicted an additional annual mortality of 1.51% combined for the three wind farms (0.60 collisions of Montagu’s Harrier males per year). An additional mortality of 4.89% (1.96 collisions per year) was predicted when applying our observed avoidance rate of 93.5% (Fig. 6).
The two different repowering scenarios yielded strongly differing results concerning the expected number of collisions: in the ‘Repowering low’ sce-nario (scesce-nario 1), mortality more than doubled compared with the original state, whereas it increased only slightly in ‘Repowering high’ (sce-nario 2; Fig. 6). The number of collisions per unit power capacity increased marginally with ‘Repow-ering low’, while it decreased by as much as
Table 2. Coefficients of binomial generalized linear models modelling the proportion of GPS fixes of Montagu's Harrier males inside wind farms as a function of the nest-to-wind-farm distance for each of the three wind farms which were regularly visited.
Delfzijl Zuid Bunde Rhede
(Intercept) 0.79 (0.87, 0.72) 2.55 (2.65, 2.44) 0.17 (1.45, 1.09)
Nest-to-wind-farm distance 1.09 (1.16, 1.02) 1.31 (1.43, 1.19) 0.99 (1.21, 0.77) Coefficients are given as means with 95% credible intervals (significant effects displayed in bold).
42.2% with ‘Repowering high’ (see Table S8 in Appendix S2 for details).
Moreover, the estimated annual mortality brought by a hypothetical new wind farm (sce-nario 3) differed greatly between locations across the study area (Fig. 7), ranging from virtually zero to 5.95% and 19.33%, for 98% and 93.5% avoid-ance, respectively (Table S8 in Appendix S2). Hence, in the‘worst case’ concerning the siting of the new wind farm, the total additional annual mortality for this scenario was considerably higher than in the present state (Fig. 6).
DISCUS SION Flight activity
Montagu’s Harriers were found to spend a large portion of the day inflight (8.2 h), which corrobo-rates earlier observations (Grajetzky & Nehls 2017, Schlaich et al. 2017). This particularly high flight activity is probably not attained by any other Central European bird of prey: e.g. 4–5 h/day in Common Kestrels Falco tinnunculus (Masman et al. 1988), 3.75 h/day in Red Kites Milvus milvus (H€otker et al. 2017) and 4–10% of daytime in Common Buzzards Buteo buteo (van Gasteren et al. 2014). It is also worth mentioning that the closely related Hen Harrier Circus cyaneus, which occurs sympatrically with Montagu’s Harrier in our study area, was shown to spend distinctly less time in flight (4–6 h/day for males during the breeding season; Dutch Montagu’s Harrier Foun-dation unpubl. data; O’Halloran in Furness et al. 2016). The high flight activity in Montagu’s Harri-ers implies an increased collision risk with wind turbines relative to other species.
Horizontal overlap with wind farms As expected, the proportion of time spent in wind farms decreased with increasing nest-to-wind-farm distance. However, wind farms were still visited to some extent even when nest-sites were as far as 5 km away, reflecting the large home-ranges of male Montagu’s Harriers during the breeding
Figure 3. Frequency distribution offlight altitude above ground level for the tracked Montagu's Harrier males in bins of 10 m. Flight altitudes above 200 m (4.0%) and below–10 m (1.4%) are not shown. Filled background rectangle: average rotor height range of wind turbines in the study area (45–125 m). [Col-ourfigure can be viewed at wileyonlinelibrary.com]
Figure 4. Proportion of GPSfixes of Montagu's Harrier males within the rotor height range of wind turbines as a function of the minimum rotor tip height above ground level. Each curve represents a given rotor radius. Circles and broken lines: tur-bine models set in place in the Bunde wind farm before (open circle) and after repowering (filled circle; Table S1 in Appendix S1).
Table 3. Observed and expected proportions of GPSfixes of Montagu's Harrier males within the risk distance from wind tur-bine towers (RD; rotor radius plus a buffer of 5 m).
Height range No. of fixes (obs. data) Obs. proportion within RD (%) Exp. proportion within RD (mean) (%) Avoidance rate (%) P-value Below RHR 7207 4.54 6.10 25.56 0.004 Within RHR 438 0.46 7.04 93.52 0.001 Above RHR 412 7.28 6.69 –8.87 0.627
Obs., observed; exp., expected; RHR, rotor height range. The P-value represents the hypothesis of the observed propor-tion being smaller than the expected (significant avoidance rates displayed in bold).
season (Schipper 1977, Trierweiler 2010). Consid-erable differences in the magnitude of this effect were found between the three wind farms, which may in part be explained by the size of the wind farm areas (Table S1 in Appendix S1). Moreover, it can be expected that habitat quality is a key aspect determining the extent to which a wind farm is visited. This offers promising opportunities for mitigation measures in the context of landscape planning: making wind farm areas unattractive both for breeding and for foraging– while offering attractive habitat outside the wind farms – will probably reduce the amount of time the birds spend inside risk areas (Grajetzky & Nehls 2017). Vertical overlap with wind turbine rotors In accordance with earlier studies (Schipper 1977, Grajetzky & Nehls 2017), we found that
Montagu’s Harriers mostly flew a few metres above the ground. Consequently, only a small frac-tion of time was spent within the height range of wind turbine rotors. For other raptors, including particularly collision-prone species such as White-tailed Eagles Haliaeetus albicilla and Red Kites, considerably higher proportions of flights at higher altitude have been shown (Averyet al. 2011, Dahl et al. 2013, H€otker et al. 2017), suggesting a rela-tively low collision risk in Montagu’s Harriers. However, the small proportion of time spent within the rotor height range is partly counter-acted by the high overall flight activity of Mon-tagu’s Harriers.
The predominantly low flight altitude implies that raising the rotor of wind turbines decreases the proportion of flights at risk altitude, which has also been demonstrated for a range of seabird spe-cies (Johnston et al. 2014). Also from the perspec-tive of the wind power industry, it is often desirable to raise the rotors of wind turbines due to increased wind speeds at higher altitudes (Hau 2006). However, reducing the distance between rotor and ground may be economically desirable in situations when the maximum height of the rotors is regulated, for example to reduce the impact on local residents regarding shadow cast and landscape aesthetics, as happened in the case of Bunde. It must be noted that although using wind turbines with a large distance between rotor and ground is probably beneficial for locally breed-ing and foragbreed-ing birds (Corman & Garthe 2014, Johnston et al. 2014, Ross-Smith et al. 2016, H€otker et al. 2017), birds commuting or migrating through the area as well as bats may face an increased collision risk with higher turbines (Bar-clay et al. 2007, Krijgsveld et al. 2009, Stumpf et al. 2011).
Avoidance behaviour
With our novel null model approach, we demon-strated that the observedflight trajectories of Mon-tagu’s Harriers inside wind farms deviated from random flight movements with respect to the dis-tance to wind turbine towers, with a distinct reduction of flight activity near the turbines. This indicates that Montagu’s Harriers adjust flight movements in response to turbines. Hence, the present study represents one of the few published cases where avoidance behaviour in birds within wind farms has been empirically assessed and
Figure 5. Avoidance behaviour of two Montagu's Harrier males towards wind turbines as derived from high-resolution GPS tracking (distance bin width: 20 m). The columns of graphs represent the tracked individuals, and the rows repre-sent the different height classes in relation to the rotor height range (RHR; 45–125 m above ground level). Negative index values of the avoidance/attraction index indicate avoidance, while positive values indicate attraction. Points: mean; grey bars: range between 2.5 and 97.5% quantiles; stars: significant bins; broken lines: rotor radius; lower right corner: sample sizes for GPSfixes and tracks. [Colour figure can be viewed at wileyonlinelibrary.com]
demonstrated based on tracking of flight move-ments (Desholm & Kahlert 2005, Plonczkier & Simms 2012, Hull & Muir 2013, Thaxter et al. 2018). Our null model approach based on simu-lated random flight paths can be applied across species and environments wherever available track-ing data have sufficient positional accuracy.
We emphasize that we only considered avoid-ance behaviour at a meso-scale (sensu Cook et al. 2014) or, in other words, avoidance of the tur-bines inside the wind farm. Neither ‘macro-avoid-ance’ (avoidance of the entire wind farm area) nor ‘micro-avoidance’ (avoidance of the sweeping rotor blade of a specific turbine) were included in our avoidance rate estimates. However, in the CRMs we implicitly took macro-avoidance into account when determining the amount of time that the tracked birds spent inside the wind farms. Conse-quently, the total avoidance rate for Montagu’s Harriers, including avoidance at all three spatial scales, may be higher. However, previous studies have suggested that most avoidance takes place at the macro- and meso-scale (Cooket al. 2014), and our results on horizontal overlap do not suggest pronounced macro-avoidance in Montagu’s Harri-ers, which accords with earlierfindings (Madden & Porter 2007, Hernandez-Pliego et al. 2015).
The meso-avoidance rate estimate for Mon-tagu’s Harriers presented here is, to our knowl-edge, the first to be published for this species.
Although the outcome (93.5%) fits relatively well within the range of documented (total) avoidance rates for other species (SNH 2016), further data collection and more detailed analyses are highly desirable to assess whether it is indeed generaliz-able and to what extent it is supplemented by micro-avoidance. In general, we note that we have tended to underestimate avoidance behaviour, as we did not account for turbine operating times (Hull & Muir 2013), and Montagu’s Harriers might have closely approached the turbines at times when these were inoperative. Moreover, our analysis of avoidance behaviour could be refined by considering the discoidal shape of the actual risk volume (whose orientation is continuously adjusted to the wind direction), instead of apply-ing a simplified cylindrical risk volume around the turbine tower. Another important issue is that we probably underestimated avoidance due to the properties of the GPS tracking system used: data from the trackers only entered the database when the birds made contact with antennas, which were usually situated close to the nest-site and did not cover the entire wind farm areas at all times,
Figure 6. Annual per-capita mortality through collisions with wind turbines for males of the studied population of Montagu's Harriers as estimated with‘Band’ collision risk models (com-bined for all wind farms). Considered scenarios: repowering of the wind farms with either low (‘Repower low’) or high modern turbines (‘Repower high’) and construction of a new wind farm (‘New WF’; ‘worst case’ concerning the location of new wind
farm shown). AR, avoidance rate. Figure 7. Sensitivity map of the Montagu's Harrier population in the study area towards the erection of a new wind farm. Dots indicate the amount of expected additional annual per-capita mortality through turbine collisions for males in a wind farm erected at the respective location as calculated with‘Band’ colli-sion risk models (98% avoidance rate; see text). Only locations with at least 0.2% mortality are displayed. WF, wind farm. [Col-ourfigure can be viewed at wileyonlinelibrary.com]
meaning that wind turbine collisions could have stayed undetected. This highlights the added value of tracking systems which transfer data via GSM or satellite. Moreover, it would be valuable to combine tracking studies with carcass searches in order to validate the avoidance rate (and mortal-ity) estimates derived from the tracking data. Collision risk models
The repowering of wind farms with modern wind turbines has been described as offering chances for reducing adverse wildlife impacts (Marques et al. 2014). By applying two different repowering sce-narios, we demonstrated that the power capacity of wind farms may be considerably increased through repowering without enhancing the expected number of collision fatalities if turbines with a high distance between ground and rotor are applied. In contrast, choosing low-reaching tur-bines would considerably increase the collision risk of Montagu’s Harriers.
However, we found the spatial distribution of wind farms to have an even stronger effect on the estimated number of collisions because the uneven distribution of nest-sites across the study area lar-gely determines the degree to which risky areas are visited. This finding is in line with previous studies demonstrating the crucial importance of wind-farm siting for the impact on bird popula-tions (Schaub 2012).
We emphasize that the absolute results from the CRMs should not be taken literally but should rather be interpreted as indications. Among other reasons, this is due to the fact that our avoidance rate estimate – which has a critical impact on the overall outcome of CRMs (Chamberlain et al. 2006) – should be regarded as preliminary (see above). Moreover, we made several simplifications when analysing the effect of the siting of a new wind farm. The predictions could be improved by (1) taking the habitat suitability of the hypotheti-cal wind farm site into account and (2) considering that the proportion of time spent at risk altitude depends on the distance to the nest-site (Appendix S4; Grajetzky & Nehls 2017).
As we have focused on male Montagu’s Harriers throughout our study, it is unclear to what extent females are exposed to wind turbine collision risk. It is likely that females have a lower collision risk than males during the breeding season, as in gen-eral they fly less (Schlaich et al. 2017) and
perform fewer display flights (Arroyo et al. 2013). Clearly, however, information on female collision risk is needed to evaluate the effects of wind tur-bine collisions at a population level. The same holds true for thorough population modelling, without which it is difficult to assess the effect of the estimated additional mortality on the popula-tion. At the present state, however, it seems unli-kely that the mortality induced by wind turbine collisions is severely detrimental at a population level, as there was no particular overall population decline observed over the last few years that could be attributed to the impact of wind farms (Dutch Montagu’s Harrier Foundation pers. comm.). In contrast, in the worst case of the ‘New wind farm’ scenario (Fig. 6), it is highly likely that the high amount of additional mortality would indeed lead to an overall population decline (additional annual male mortality of 7–24% compared with an esti-mated overall background mortality of 21% for adults; Bourrioux et al. 2017). This highlights the need for careful selection of new wind farm loca-tions and keeping Montagu’s Harriers’ core breed-ing areas free from such developments.
Conclusions and management implications
Based on an extensive GPS tracking dataset, we investigated the different aspects of flight beha-viour related to wind turbine collision risk in Mon-tagu’s Harriers, including meso-scale avoidance behaviour. This provided a more complete and more accurate understanding of collision risk than provided in previous studies based on visual obser-vations. It is important to stress that parts of our analyses could only be conducted with tracking data collected at high temporal resolution, as these are the only GPS tracking data that have sufficient positional accuracy. The latter is important for any researcher interested in flight altitude and avoid-ance behaviour.
Our study has two main management implica-tions:
1 Combining in-depth knowledge on collision-re-lated aspects of flight behaviour with data on the distribution of nest-sites allows estimation of the expected number of collisions for differ-ent scenarios of wind farm siting and design. This approach potentially represents an advan-tageous alternative to local environmental
assessment studies for individual wind farm development sites.
2 Our results indicate that wind turbine collision risk for harriers could be reduced either by using turbines with a larger distance between rotor and ground or by constructing wind farms at larger distances from nest-sites. However, as the spatial distribution of wind farms has a much stronger effect on the estimated number of collisions compared with the design of the turbines, precluding wind energy developments from core breeding areas remains the most important mitigation measure.
In conclusion, we demonstrate the great poten-tial of detailed GPS tracking data for assessing wind turbine collision risk in birds. It would be valuable to expand our approach to other species with different flight characteristics, which may require different mitigation measures.
We thank the staff and volunteers of the Dutch Mon-tagu’s Harrier Foundation for fieldwork; J. Eccard and the Animal Ecology Group at the University of Pots-dam/Germany for co-supervising T.S.; S. Schirmer for advice on statistics; the farmers in East Groningen and Rheiderland for good collaboration; the wind farm oper-ators for technical information; A. Villers, A. Johnston, A. Brenninkmeijer, C. Grande, H. Illner, K. Etling, O. Krone and V. Ross-Smith for various support; Chris Thaxter, Beatriz Arroyo, Jeremy Wilson and two anony-mous reviewers for helpful comments on earlier versions of the manuscript; and the Dutch Ministry of Agricul-ture, Nature and Food Quality, the Province of Gronin-gen, Prins Bernhard Cultuurfonds and B.V. Oldambt for funding. TheUvA-BiTS infrastructure was facilitated by Infrastructures for E-Science, developed with the sup-port of the Netherlands eScience Centre (NLeSC) and LifeWatch, and conducted on the Dutch National E-Infrastructure with support from the SURF Foundation. GPS tracking was approved by the responsible authori-ties: University of Groningen (permits 5869B and 6429B) and LAVES Niedersachsen (permit 33.12-42502-04-14/1550).
AUTHOR CONT RIBUTIONS
B.K., R.K. and A.S. initiated the study. T.S. and R.K. designed the study concept. B.K. acquired funding. R.K., A.S., T.S. and B.K. did fieldwork. W.B. provided support for data acquisition and processing. T.S. processed the data, conducted the statistical analyses and wrote the first draft. All authors contributed critically to subsequent drafts and gavefinal approval for publication.
D AT A AV AIL AB ILITY ST AT EME NT The data that support the findings of this study are available from the corresponding author upon reasonable request.
REFERENCES
Arroyo, B., Mougeot, F. & Bretagnolle, V. 2013. Characteristics and sexual functions of sky-dancing displays in a semi-colonial raptor, the Montagu's Harrier (Circus pygargus). J. Raptor Res. 47: 185–196.
Avery, M.L., Humphrey, J.S., Daughtery, T.S., Fischer, J.W., Milleson, M.P., Tillman, E.A., Bruce, W.E. & Walter, W.D. 2011. Vulture flight behavior and implications for aircraft safety. J. Wildl. Manage. 75: 1581–1587.
Band, W. 2017. Calculation of collision risk for bird passing through rotor area. Available at: http://www.snh.gov.uk/docs/ C234672.xls (accessed 1 March 2017).
Band, W., Madders, M. & Whitfield, D.P. 2007. Developing field and analytical methods to assess avian collision risk at wind farms. In de Lucas, M., Janns, G.F.E. & Ferrer, M. (eds) Birds and Wind Farms: Risk Assessment and Mitigation: 259–275. Madrid: Quercus.
Barclay, R.M.R., Baerwald, E.F. & Gruver, J.C. 2007. Variation in bat and bird fatalities at wind energy facilities: assessing the effects of rotor size and tower height. Can. J. Zool. 85: 381–387.
Barrios, L. & Rodrıguez, A. 2004. Behavioural and environmental correlates of soaring-bird mortality at on-shore wind turbines. J. Appl. Ecol. 41: 72–81.
Bellebaum, J., Korner-Nievergelt, F., D€urr, T. & Mammen, U. 2013. Wind turbine fatalities approach a level of concern in a raptor population. J. Nat. Conserv. 21: 394–400. Boessenkool, B. 2017. OSMscale: Add a scale bar to
‘OpenStreetMap’ plots. R package version 0.4.1.
Bourrioux, J., Printemps, T., van Hecke, B., Villers, A., Chadœuf, J., Augiron, S., Bretagnolle, V., Millon, A. & le reseau busards. 2017. Ten years after: synthesis of the national tagging scheme on Montagu's Harriers in France. Ornithos 24: 305–322.
Bouten, W., Baaij, E.W., Shamoun-Baranes, J. & Camphuysen, C.J. 2013. A flexible GPS tracking system for studying bird behaviour at multiple scales. J. Ornithol. 154: 571–580.
Calenge, C. 2015. Analysis of animal movements in R: The adehabitatLT package. Available at: https://cran.r-project.org/ web/packages/adehabitatLT/vignettes/adehabitatLT.pdf (accessed 30 April 2019).
Chamberlain, D.E., Rehfisch, M.R., Fox, A.D., Desholm, M. & Anthony, S.J. 2006. The effect of avoidance rates on bird mortality predictions made by wind turbine collision risk models. Ibis 148: 198–202.
Cook, A.S.C.P., Humphreys, E.M., Masden, E.A. & Burton, N.H.K. 2014. The avoidance rates of collision between birds and offshore turbines. Scot. Marine Freshw. Sci. 5: 1–247. Corman, A.-M. & Garthe, S. 2014. Whatflight heights tell us
about foraging and potential conflicts with wind farms: a case study in Lesser Black-backed Gulls (Larus fuscus). J. Ornithol. 155: 1037–1043.
Dahl, E.L., Bevanger, K., Nygard, T., Røskaft, E. & Stokke, B.G. 2012. Reduced breeding success in White-tailed Eagles at Smøla windfarm, western Norway, is caused by mortality and displacement. Biol. Conserv. 145: 79–85. Dahl, E.L., May, R., Hoel, P.L., Bevanger, K., Pedersen,
H.C., Røskaft, E. & Stokke, B.G. 2013. White-tailed Eagles (Haliaeetus albicilla) at the Smøla wind-power plant, central Norway, lack behavioral flight responses to wind turbines. Wildl. Soc. Bull. 37: 66–74.
D'Eon, R.G. & Delparte, D. 2005. Effects of radio-collar position and orientation on GPS radio-collar performance, and the implications of PDOP in data screening. J. Appl. Ecol. 42: 383–388.
Desholm, M. & Kahlert, J. 2005. Avian collision risk at an offshore wind farm. Biol. Lett. 1: 296–298.
Dray, S. & Dufour, A.B. 2007. The ade4 package: implementing the duality diagram for ecologists. J. Stat. Softw. 22: 1–20.
D€urr, T. 2017. Bird fatalities at wind turbines in Europe. Buckow b. Nennhausen: Vogelschutzwarte Brandenburg. Available at: http://www.lfu.brandenburg.de/cms/media.php/ lbm1.a.3310.de/wka_voegel_eu.xls (accessed 23 April 2017).
Fellows, I. 2016. OpenStreetMap: Access to Open Street Map raster images. R package version 0.3.3.
Furness, R.W., Trinder, M., MacArthur, D. & Douse, A. 2016. A theoretical approach to estimating bird risk of collision with wind turbines where empirical flight activity data are lacking. Energy Power Eng. 8: 183–194.
van Gasteren, H., Both, I., Shamoun-Baranes, J., Lalo€e, J.-O. & Bouten, W. 2014. A pilot study on GPS tracking of Common Buzzards Buteo buteo at military airfields to advance bird strike prevention. Limosa 87: 107–116. Gelman, A. & Su, Y.-S. 2015. arm: Data analysis using
regression and multilevel/hierarchical models. R package version 1.8-6.
Grajetzky, B. & Nehls, G. 2017. Telemetric monitoring of Montagu's Harrier in Schleswig-Holstein. In H€otker, H., Krone, O. & Nehls, G. (eds) Birds of Prey and Wind Farms. Analysis of problems and possible solutions: 97–148. Berlin: Springer.
Hau, E. 2006. Wind Turbines. Fundamentals, Technologies, Application, Economics, 2nd edn. Berlin: Springer.
Hernandez-Pliego, J., de Lucas, M., Mu~noz, A.-R. & Ferrer, M. 2015. Effects of wind farms on Montagu's Harrier (Circus pygargus) in southern Spain. Biol. Conserv. 191: 452–458. H€otker, H., Mammen, K., Mammen, U. & Rasran, L. 2017.
Red Kites and wind farms – Telemetry data from the core breeding range. In K€oppel, J. (ed.) Wind Energy and Wildlife Interactions. Presentations from the CWW2015 Conference: 3–15. Berlin: Springer.
Hull, C.L. & Muir, S.C. 2013. Behavior and turbine avoidance rates of eagles at two wind farms in Tasmania, Australia. Wildl. Soc. Bull. 37: 49–58.
Jarvis, A., Reuter, H.I., Nelson, A. & Guevara, E. 2008. Hole-filled Seamless SRTM Data V4. Palmira: International Centre for Tropical Agriculture. Available at: http://srtm.csi.c giar.org (accessed 30 April 2019).
Johnston, A., Cook, A.S.C.P., Wright, L.J., Humphreys, E.M. & Burton, N.H.K. 2014. Modelling flight heights of marine birds to more accurately assess collision risk with offshore wind turbines. J. Appl. Ecol. 51: 31–41.
Kenward, R. 1987. Wildlife Radio Tagging. Equipment, Field Techniques and Data Analysis. London: Academic Press. Korner-Nievergelt, F., Roth, T., von Felten, S., Guelat, J.,
Almasi, B. & Korner-Nievergelt, P. 2015. Bayesian Data Analysis in Ecology using Linear Models with R, BUGS, and Stan. Amsterdam: Academic Press.
Krijgsveld, K.L., Akershoek, K., Schenk, F., Dijk, F. & Dirksen, S. 2009. Collision risk of birds with modern large wind turbines. Ardea 97: 357–366.
Langley, R.B. 1999. Dilution of precision. GPS World 10: 52–59.
Madden, B. & Porter, B. 2007. Do wind turbines displace Hen Harriers Circus cyaneus from foraging habitat? Preliminary results of a case study at the Derrybrien Wind Farm, County Galway. Ir. Birds 8: 231–236.
Madders, M. & Whitfield, D.P. 2006. Upland raptors and the assessment of windfarm impacts. Ibis 148: 43–56.
Marques, A.T., Batalha, H., Rodrigues, S., Costa, H., Pereira, M.J.R., Fonseca, C., Mascarenhas, M. & Bernardino, J. 2014. Understanding bird collisions at wind farms: an updated review on the causes and possible mitigation strategies. Biol. Conserv. 179: 40–52.
Masman, D., Daan, S. & Dijkstra, C. 1988. Time allocation in the Kestrel (Falco tinnunculus), and the principle of energy minimization. J. Anim. Ecol. 57: 411–432.
Pebesma, E.J. & Bivand, R.S. 2005. Classes and methods for spatial data in R. R News 5: 9–13.
Peron, G., Fleming, C.H., Duriez, O., Fluhr, J., Itty, C., Lambertucci, S., Safi, K., Shepard, E.L.C. & Calabrese, J.M. 2017. The energy landscape predictsflight height and wind turbine collision hazard in three species of large soaring raptor. J. Appl. Ecol. 54: 1895–1906.
Plonczkier, P. & Simms, I.C. 2012. Radar monitoring of migrating Pink-footed Geese: behavioural responses to offshore wind farm development. J. Appl. Ecol. 49: 1187– 1194.
R Core Team 2017. R: A Language and Environment for Statistical Computing. Vienna: R Foundation for Statistical Computing.
Richard, E., Calenge, C., Sa€ıd, S., Hamann, J.-L. & Gaillard, J.-M. 2012. Studying spatial interactions between sympatric populations of large herbivores: a null model approach. Ecography 36: 157–165.
Ross-Smith, V.H., Thaxter, C.B., Masden, E.A., Shamoun-Baranes, J., Burton, N.H.K., Wright, L.J., Rehfisch, M.M., Johnston, A. & Thompson, D. 2016. Modelling flight heights of Lesser Black-backed Gulls and Great Skuas from GPS: a Bayesian approach. J. Appl. Ecol. 53: 1676–1685. Schaub, M. 2012. Spatial distribution of wind turbines is
crucial for the survival of Red Kite populations. Biol. Conserv. 155: 111–118.
Schipper, W.J.A. 1977. Hunting in three European harriers (Circus) during the breeding season. Ardea 65: 53–72. Schlaich, A.E., Klaassen, R.H.G., Bouten, W., Both, C. &
Koks, B.J. 2015. Testing a novel agri-environment scheme based on the ecology of the target species, Montagu's Harrier Circus pygargus. Ibis 157: 713–721.
Schlaich, A.E., Bouten, W., Bretagnolle, V., Heldbjerg, H., Klaassen, R.H.G., Sørensen, I.H., Villers, A. & Both, C. 2017. A circannual perspective on daily and total flight distances in a long-distance migratory raptor, the Montagu's Harrier, Circus pygargus. Biol. Lett. 13: 20170073.
Schuster, E., Bulling, L. & K€oppel, J. 2015. Consolidating the state of knowledge: a synoptical review of wind energy's wildlife effects. Environ. Manage. 56: 300–331.
SNH 2016. Avoidance rates for the onshore SNH wind farm collision risk model. Available at: http://www.snh.gov.uk/doc s/A2126483.pdf (accessed 1 March 2017).
Stumpf, J.P., Denis, N., Hamer, T.E., Johnson, G. & Verschuyl, J. 2011. Flight height distribution and collision risk of the Marbled Murrelet Brachyramphus marmoratus: methodology and preliminary results. Mar. Ornithol. 39: 123– 128.
Tabassum-Abbasi, P.M., Abbasi, T. & Abbasi, S.A. 2014. Wind energy: increasing deployment, rising environmental concerns. Renew. Sustain. Energy Rev. 31: 270–288. Thaxter, C.B., Ross-Smith, V.H., Bouten, W., Masden, E.A.,
Clark, N.A., Conway, G.J., Barber, L., Clewley, G. & Burton, N.H.K. 2018. Dodging the blades: new insights into three-dimensional area use of offshore wind farms by Lesser Black-backed Gulls Larus fuscus. Inter-Res. Mar. Ecol. Prog. Ser. 587: 247–253.
Trierweiler, C. 2010. Travels to feed and food to breed. The annual cycle of a migratory raptor, Montagu's Harrier, in a modern world. PhD Thesis. University of Groningen, The Netherlands. Available at: http://www.rug.nl/research/portal/ en/publications/travels-to-feed-and-food-to-breed(adc4194b-e527-45c5-ae8f-55ba8b095fd1).html (accessed 30 April 2019).
Vasilakis, D.P., Whitfield, D.P., Schindler, S., Poirazidis, K.S. & Kati, V. 2016. Reconciling endangered species conservation with wind farm development: Cinereous Vultures (Aegypius monachus) in south-eastern Europe. Biol. Conserv. 196: 10–17.
Received 3 December 2018; revision accepted 29 September 2019.
Associate Editor: Beatriz Arroyo.
SUPPORTING INFORMATION
Additional supporting information may be found online in the Supporting Information section at the end of the article.
Appendix S1. Supplementary information on methods (including Tables S1–S7 and Figures S1–S2).
Appendix S2. Supplementary results (including Table S8 and Figures S3–S4).
Appendix S3. Supplementary analysis of verti-cal precision (including Table S9).
Appendix S4. Supplementary analysis of verti-cal overlap in relation to different predictor vari-ables (including Table S10).
