Edited by G.C. Boere, C.A. Galbraith and D.A. Stroud
Assisted by L.K. Bridge, I. Colquhoun, D.A. Scott,
D.B.A. Thompson and L.G. Underhill

Names used for geographical entities do not imply recognition, by the organisers of the Waterbirds around the world conference or other supporting organisations or governments, of the political status or boundaries of any particular territory. Names of territories used (and any alternatives) are included solely to help users of this publication apply information contained within this volume for waterbird conservation purposes. The views expressed in papers included within this volume do not necessarily represent views of the editors or the organisations and governments that supported the conference and this publication.
Whooper Swans Cygnus cygnus arriving at Martin Mere, England. Photo: Paul Marshall.
(www.paulmarshallphotography.com)
Copyright of all photographs used in this publication resides with the named photographers.

Waterbirds around the world
Anthony D. Fox & Ib Krag Petersen
Department of Wildlife Ecology and Biodiversity, National Environmental Research Institute, Kalø, Grenåvej 14, DK-8410 Rønde,
Denmark. (email: tfo@dmu.dk)
Fox, A.D. & Petersen, I.K. 2006. Assessing the degree of habitat loss to marine birds from the development of offshore wind farms.
Waterbirds around the world. Eds. G.C. Boere, C.A. Galbraith & D.A. Stroud. The Stationery Office, Edinburgh, UK. pp. 801-804.
ABSTRACT
Environmental impact assessment of offshore wind farms requires an assessment of waterbird habitat loss, both physical and as a result of behavioural avoidance. This paper briefly summarizes the aerial survey transect sampling methods and approach adopted to undertake such assessments at two marine wind farms constructed in Danish waters in 2003. At present, bird encounter rates per unit effort on track have been used to assess the “preferences” of bird species for wind farm and adjacent areas and to make a statistical comparison of abundance and distribution before and after construction. Preliminary results from the two case studies are presented. These results suggest that divers (Gaviidae), grebes (Podicipedidae), gannets
(Sulidae), some seaducks (Anatidae) and auks (Alcidae) avoid wind farms after the erection of turbines, whilst some species of gulls (Laridae) and terns (Sternidae) show some preference for sites in which wind farms have been constructed compared to the previously undeveloped sites. A description of an improved application is given. This combines distance sampling and spatial modelling techniques currently being developed, and will provide more robust comparisons of waterbird density surfaces with respect to pre and post-construction development scenarios to support more effective environmental impact assessment in marine environments.
turbines because of the confining visual effects, such that even at avoidance distances less than half the distance between turbines, net habitat loss equivalent to the size of the wind farm occurs.
Since birds may habituate over time to unfamiliar constructions in their foraging distribution at sea, there may be a temporal component to their avoidance responses.
Environmental impact assessment of offshore wind farms requires a basic evaluation of the significance of such habitat loss. Indeed, assessment of the consequences at local and population/flyway level may be a foregone requirement, for example, under environmental criteria placed on construction. Hence, it is important to quantify habitat loss in a way that accounts for the spatial and temporal heterogeneity in bird distributions prior to and after construction. In this short account, we summarize the approach taken to assess effective habitat loss to foraging waterbirds based on experiences from two wind farms constructed in
Danish inshore waters mainly during 2003. These are situated at
Horns Rev (off the exposed west coast of Jutland in the North
Sea) and Nysted (south of Rødsand in the brackish Baltic). Both projects have already been widely reported elsewhere (e.g.
Christensen et al . 2004, Kahlert et al . 2004, Petersen et al . 2004), but are summarized here. We also discuss the current state of development of the necessary tools required for effective assessment of the degree of habitat loss to foraging birds offshore.
INTRODUCTION
The dramatic development of offshore wind farms in inshore/nearshore marine waters around Europe in recent years has focussed attention on the hazards they present to waterbirds and other migratory birds that encounter these constructions.
Direct impacts, such as result from direct collision, cause mortality that affect the demography of populations, but local effects (such as the extra energy costs incurred by avoidance flights or habitat loss) are more difficult to quantify. Habitat gain and loss as a result of the construction of wind turbines can be considered on two spatial and temporal scales. Firstly, long-term but small-scale change results from the physical loss of feeding substrate under foundations and anti-scour protection and the physical gain associated with the creation of these new substrates. However, since the area of foundations and anti-scour protection rarely exceeds 2% of the total sea area covered by a wind farm, this change in food availability is considered trivial in most instances. Secondly, effective habitat loss may result from behavioural displacement of foraging birds, the response to specific stimuli such as rotating turbines and/or the activity of maintenance vessels in the vicinity. In theory, if birds avoid coming closer that half the distance between adjacent turbines, the effective loss of habitat exceeds the entire area of the wind farm. In practice, birds may avoid going between the rows of
METHODS
The strategic approach adopted in the study of the effects of the two Danish offshore wind farms has been to use aerial surveys to describe changes in bird abundance and distribution as a proxy measure of habitat loss. The problem associated with such assessments is the very large degree of spatial and temporal heterogeneity associated with bird distributions in dynamic media such as inshore marine waters. Benthic feeders (such as eiders Somateria spp. and scoters Melanitta spp. feeding on marine bivalve molluscs) may show relatively simple responses to factors such as water depth and substrate type that determine the nature and profitability of the benthos they feed upon.
Nevertheless, temporal variation in the abundance of benthic feeders will be subject to temporal variation in spat settlement and age class distributions of their prey. Bird species preying on pelagic fish, such as divers (Gaviidae), grebes (Podicipedidae) and auks (Alcidae), are even less predictable, dependent upon temporal and spatial patterns of distribution and abundance of their highly mobile prey.
To enable a pre and post-construction comparison of the bird distribution and abundance in such heterogeneous systems, it is essential to survey and re-survey a sufficiently large geographical area with high temporal frequency. This BACI (before-after control-impact) type of design should encompass sampling bird
801

Waterbirds around the world
Fig. 1.
Diagrammatic representation of theoretically constant bird densities along a transect prior to construction of offshore wind farms
(broken line), compared with post construction (solid line). The features of interest are the extent of complete habitat loss (solid arrow) and the extent of reduced densities (broken arrow).
Fig. 2.
Diagrammatic representation of theoretical modelled bird densities (with confidence intervals) along a transect sampled prior to construction of offshore wind farms (pecked line), compared with post construction (solid line). The features of interest are the extent of complete habitat loss (solid arrow), the extent of reduced densities
(pecked arrow), and hence the difference in overall densities between the two samples.
densities within:
• the area physically affected by the construction area;
• an area around this where behavioural avoidance may also cause an effect (assumed to be a gradient of avoidance with increasing distance from the turbines); and
• a reference area where bird distributions are likely to be unaffected by the construction of the wind turbines.
Such sampling needs to be undertaken as frequently as necessary to characterize changes in temporal abundance throughout the annual cycle. In two-dimensional space, we may hypothesize complete bird avoidance within a specific distance of the turbines (Fig. 1). In this case, the objective is to measure the area of displaced birds, both in terms of defining the extent of areas of water without birds and areas of water with reduced densities relative to the baseline. The ideal objective for any sampling protocol is therefore to generate bird density surfaces over large areas of open sea using data:
• from as large a reference area as possible;
• sampled as simultaneously as possible;
• gathered with the greatest level of spatial precision possible;and
• using observation platforms that cause least disruption to the patterns of abundance and distribution of undisturbed birds.
For this reason, the Danish studies have adopted aerial survey using high-winged twin-engined aircraft to cover large areas of open marine waters as rapidly as possible using internationally agreed standardized data collection protocols (described in Camphuysen et al . 2004).
Initially, a very simple analytical approach has been adopted to analyse these transect count data, based on encounter rates of birds or bird clusters observed on transects per unit effort. This approach determines the relative number of birds of different species that would be susceptible to potential disturbance effects from the wind turbines based on encounter rate corrected for observation effort. The method assesses the relative importance of the wind farm area and the adjacent waters, generally a zone within 2 and 4 km of the outermost turbines of a wind farm
(see methods in Petersen et al . 2004). The method describes the
“preference” of bird species for the wind farm area and adjacent zones of differing extent immediately adjacent to the construction area. In doing so, we define the potential impact of loss of these areas relative to the preference shown by the species for the entire study area. In a typical analysis, species encounter rate was calculated for all pre-construction surveys combined, and compared with the data generated from post-construction surveys. For these zones, the preference of the most numerously occurring species is calculated using Jacob’s selectivity index
(Jacobs 1974). Jacob’s selectivity index ( D ) varies between
-1 (all birds present outside the area of interest) and +1 (all birds inside the area of interest), and is calculated as:
D = (r - p)/ (r + p - 2rp) where r = the proportion of birds in the area of interest compared to the birds in the whole study area, and p = the proportion of the survey effort in the area of interest compared to the total survey effort in the whole study area. The difference between the two proportions is tested as the difference between the observed number of birds in the area of interest and the number expected in this area, estimated from the share of the survey effort in relation to survey effort in the total area (one-sample
χ
2 -test). Tests are made on the basis of number of observed clusters, rather than birds, because observations of individual birds fail to meet the statistical criteria of being independent. However, for some species a cluster can represent a wide range of number of individual birds, varying from 1 to 26 000 in the case of the Common
Scoter Melanitta nigra , so the use of cluster data may appear unhelpful.
An alternative approach has been to compare the pre- and post-construction frequency distribution of birds at increasing distance intervals out from wind turbines. Using GIS and other tools to construct species cumulative percentage frequency distributions of bird numbers in successive 500 m distance intervals away from the wind turbines, it is possible to use non-parametric tests to compare pre- and post-construction distributions for significant displacement effects.
However, the ultimate aim has been to develop a more sophisticated suite of analytical tools to permit more robust comparisons of before/after densities of birds. The use of survey data collected at different distances from the observer aircraft provides transect counts of birds assigned to distance categories out from the transect track-lines. Such a line-transect count approach allows the use of Distance Sampling techniques
802

Waterbirds around the world
(Buckland et al . 2001) to generate bird densities, by modelling detectability functions to correct bird density estimates for the decline in detectability probability with increasing distance from the observer. The use of such techniques also enables the incorporation of factors and covariates (such as individual observer functions, differences in sea state and light conditions, etc.) into modelled densities of birds encountered. The ultimate objective for this project has been to develop spatial modelling techniques that use line-transect counts as samples and generalized additive models to construct bird density surfaces as a function of spatially explicit covariates (e.g. Hedley et al . 1999, Hedley &
Buckland 2004, Clarke et al . 2004). This method offers an approach to the theoretical objectives established in Fig. 1 by constructing a modelled bird density surface with confidence intervals over extended areas of sea generated from aerial linetransect census data (see Fig. 2). For benthic feeding birds, the covariates used in the predictive-modelling exercise could include environmental factors such as water depth which determine profitability of shallow inshore waters (e.g. Common
Scoters forage more in shallow water depths than would be expected by chance; see Fig. 17 in Petersen et al . 2004). In the fullness of time, other environmental covariates, such as bottom substrate type and bottom aspect can be incorporated to improve the effectiveness of such model estimates. For seabirds dependent on a more dynamic food base, such as pelagic fish, the challenges to generating density surfaces are considerably greater. However, incorporation of macro-environmental parameters, such as current, salinity and water temperature profiles
(which are now routinely modelled throughout the water column around many European coasts) which correlate with prey abundance, offers some opportunities to generate bird density surfaces for these organisms as well.
At present, software to generate such modelled density surfaces is still being developed in collaboration with the Centre for Research into Ecological and Environmental Modelling at the University of St Andrews in Scotland, for implementation in the near future.
RESULTS
Results from tests based on the numbers of bird clusters encountered have been used to compare pre-construction and postconstruction distributions to look for changes in preference for the wind-farm area and surroundings between the two “treatments”. Use of the Jacobs selectivity index has suggested changes in distribution of different species, both in terms of avoidance of, and attraction to, the structures, as outlined below
(see also Christensen et al . 2004 and Petersen et al . 2004).
At both sites, analysis of the preference indices calculated for different species prior to construction confirmed that the majority of waterbirds avoided the wind farm area at both sites.
Bird densities were generally low before any construction activities on the site, making comparisons after construction difficult, especially since at present there is only a single post-construction year available for comparisons.
At Horns Rev, divers, Northern Gannets Sula bassana ,
Common Scoters and Guillemots/Razorbills Uria aalge/Alca torda showed an increased avoidance of the wind farm area (and zones within 2 and 4 km of it) after the erection of the wind turbines. In contrast, Herring Gulls Larus argentatus showed a decreased avoidance of the wind farm area, while Great Blackbacked Gulls L. marinus , Little Gulls L. minutus and
Arctic/Common Terns Sterna paradisea/S. hirundo showed a shift from avoidance before construction to a preference for the wind farm area after construction (Petersen et al . 2004).
At Nysted, Long-tailed Ducks Clangula hyemalis and
Common Eiders Somateria mollissima showed a reduced preference for the wind farm area (and zones within 2 and 4 km of it) after the erection of the wind turbines. Herring Gulls showed a slight increase in preference for the wind farm area and the zones around it after construction (Kahlert et al . 2004).
DISCUSSION
The bird studies being carried out at Nysted and Horns Rev during the period 1999-2006 were initiated under the terms and conditions placed upon the granting of permission to construct wind farms at the two sites by the Danish authorities. These studies have been carried out before, during and after construction of both wind farms. The installation of wind turbines was finished in autumn 2002 (Horns Rev) and summer 2003 (Nysted). However, the construction phase was too short to offer any opportunity to assess the effects of the physical construction of turbines on bird distribution. Construction activities coincided with periods of the year when fewest feeding birds were in the vicinity (by prior design), with the result that no assessment could be made of the disturbance effects of building work during that phase.
Furthermore, the data reported here (for the period up to 2004) represent material gathered from one year or less into the initial operational phase of the wind farms. For this reason, it is not possible to quantify natural variation between years, seasons, species and sites and the possible habituation effects during the operational phase. Therefore, it must be emphasized that these results are to be considered as preliminary, and must await further compilation of data before firm conclusions can be drawn with respect to the impact on birds. The final environmental impact assessment for the two wind farms will be undertaken upon termination of the environmental monitoring programmes in 2006.
There remains considerable scope to improve on these methods and to test many of the assumptions associated with their use. In particular, distance sampling necessitates that all objects on the track-line are observed from the count platform, an assumption that has considerable bearing on the ability to generate unbiased density estimates. Studies to assess whether this assumption is met, using double-platform repeat counting, have been attempted, including the use of two aircraft and vertical photographic techniques to capture scenes of the distribution of birds on the sea surface to which count observers were exposed in the following aircraft. Such techniques proved so effective, and the imagery quality sufficient to enable automated computer identification and counting of bird “objects” on the photographs, that alternative methods of counting birds at sea have presented themselves. The use of geo-rectified vertical photography of large areas of sea surface, and subsequent computer based pattern recognition software to identify, position with great spatial accuracy and ultimately count birds on that surface, frees the need for distance sampling to generate bird densities, and offers some very exciting prospects for future developments in this field.
ACKNOWLEDGEMENTS
We gratefully acknowledge the help and support of our NERI colleagues, especially the aerial survey observers, Ib Clausager,
803

Waterbirds around the world
Johnny Kahlert, Thomas Kjær and Mark Desholm. Our thanks also to David Borchers, Sharon Hedley and colleagues at
St Andrews University for support with developing analytical techniques and to E2, Elsam and their staff for funding and support for the two Danish case studies reported here.
REFERENCES
Buckland, S.T., Anderson, D.R., Burnham, K.P., Laake, J.L.,
Borchers, D.L. & Thomas, L.
2001. Introduction to
Distance Sampling - Estimating Abundance of
Biological Populations. University Press, Oxford.
Camphuysen, C.J., Fox, A.D., Leopold, M.F. & Petersen, I.K.
2004. Towards standardised seabirds at sea census techniques in connection with environmental impact assessments for offshore wind farms in the U.K. A comparison of ship and aerial sampling methods for marine birds, and their applicability to offshore wind farm assessments.
Report to Collaborative Offshore Wind Research into the
Environment (COWRIE), Crown Estate Commissioners,
London. Available at: http://www.thecrownestate.co.uk/
1352_bird_survey_phase1_final_04_05_06.pdf.
Christensen, T.K., Hounisen, J.P., Clausager, I. & Petersen,
I.K.
2004. Visual and radar observations of birds in relation to collision risk at the Horns Rev offshore wind farm.
Annual status report 2003. Report request.
Commissioned by Elsam Engineering A/S. National
Environmental Research Institute. Available at: http://www.hornsrev.dk/Miljoeforhold/miljoerapporter/
Visual_radar_observations_2003_status_report.pdf.
Clarke, E.D., Spear, L.B., McCracken, M.L., Marques,
F.F.C., Borchers, D.L., Buckland, S.T. & Ainley, D.G.
2004. Validating the use of generalized additive models and at-sea surveys to estimate size and temporal trends of seabird populations. Journal of Applied Ecology 40:
278-292.
Hedley, S.L. & Buckland, S.T.
2004. Spatial models for line transect sampling. Journal of Agricultural, Biological and Environmental Statistics 9: 181-199.
Hedley, S.L., Buckland, S.T. & Borchers, D.L.
1999. Spatial modelling from line transect data. Journal of Cetacean
Research and Management 1: 255-264.
Jacobs, J. 1974. Quantitative measurements of food selection.
Oecologia 14: 413-417.
Kahlert, J., Petersen, I.K., Fox, A.D., Desholm, M. &
Clausager, I.
2004. Investigations of birds during construction and operation of Nysted offshore wind farm at Rødsand - Annual status report 2003. Report request.
Commissioned by Energi E2 A/S. National
Environmental Research Institute. Available at: http://uk.nystedhavmoellepark.dk/upload/pdf/Birds2003
.pdf.
Petersen, I.K., Clausager, I. & Christensen, T.K. 2004. Bird numbers and distribution in the Horns Rev offshore wind farm area. Annual status report 2003. Report commissioned by Elsam Engineering A/S. National
Environmental Research Institute. Available at: http://www.hornsrev.dk/Miljoeforhold/miljoerapporter/Bird_numbers_2003%20status_report.pdf.
The Nysted Offshore Wind Farm in the Danish part of the Baltic Sea consists of 72 turbines situated c. 11.5 km south of the island of Lolland and covering an area of 24 km 2 . Water depth in the area is 6 - 9.5 m. The windfarm has a total power capacity of 165.6 MW, and is the world's largest offshore wind farm. (Photo provided by Energy E2). Photo: Ib Krag Petersen.
804