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

3. Definitions of Avoidance Behaviour

3.1 Introduction

Chamberlain et al. (2005, 2006) demonstrated that, of the parameters used in the Band collision model (Band 2006), the avoidance rate used was among those that the predicted collision rates were most sensitive to. Subsequently, the identification of appropriate avoidance rates has been subject to widespread debate. Guidance produced by Scottish Natural Heritage ( SNH 2010) has been largely accepted in the UK for the terrestrial environment, subject to revision as additional data become available ( e.g. Pendlebury 2006). Whilst this document references some seabird species, its guidance for offshore windfarms is limited to the suggestion that a range of avoidance rates should be presented. Country agencies have provided advice to developers as necessary, but the lack of guidance produced specifically for the offshore environment, and for the updated Band model for use in the offshore environment (Band 2012), has led to uncertainty amongst developers, regulators and other stakeholders as to what values reflect realistic avoidance rates ( e.g. MacArthur Green 2012, MORL 2012) and for which collision risk models they are appropriate. Previous studies have attempted to review avoidance behaviour in offshore species ( e.g. Maclean et al. 2009, Cook et al. 2012) but a failure to gain widespread consensus about the values presented has meant the situation remains largely unresolved.

Deriving avoidance rates for terrestrial windfarm developments has been based largely on the ability to estimate the numbers of birds killed by collisions. Every bird flying through the rotor-swept area of a turbine has a probability of colliding with the turbine blades (P _coll), typically in the range of 5-10% for seabirds, depending on species and the design of the turbine concerned (Cook et al. 2011). By multiplying the total number of birds expected to pass through the rotor-swept area of a turbine by P _coll it is possible to predict the number of collisions that would be expected, should birds take no action to avoid collision. In the case of terrestrial windfarms estimates of the total number of collisions actually occurring, once turbines are operational, can be made by using corpse searches around the windfarm to assess actual mortality rates, or observed collision rates ^[1] . Band (2000) therefore suggests that the avoidance rate can be thought of as equation 6, where the collision rate expected in the absence of avoidance is the total number of birds (Flux rate) passing through the rotor-swept area of a turbine, multiplied by P _coll. However, in practice both P _coll and the flux rate are likely to be subject to error - P _coll in relation to the model input parameters and flux rate in relation to estimates of the total number of birds passing through the windfarm. Of the two, the error associated with the flux rate is likely to be greatest as a result of the difficulty in recording the number of birds passing through a site over an extended period of time and the need to extrapolate from, often brief, observation periods to estimate a flux rate for the study period as a whole. As a result of the need to incorporate this error, it may be better to think of this in terms of an avoidance correction factor, as opposed to an avoidance rate, which implies it is solely dependent on the behavioural responses of birds:

However, in the case of offshore windfarms, recording actual collisions, or mortality rates, is not currently practical, although the forthcoming Offshore Renewables Joint Industry Project ( ORJIP) will aim to provide additional data to inform avoidance rates using behavioural observations (Davies et al. 2013). Therefore, at present, guidance on appropriate avoidance rates for use in the offshore environment draws on the experiences gained in the terrestrial environment, as well as being informed by studies of bird movements, where suitable data are available ( e.g. Desholm & Kahlert 2005, Petersen et al. 2006, Masden et al. 2009, Blew et al. 2008, Krijgsveld et al. 2011). Where studies have sought to use movement data to inform values for avoidance rates, this has often led to confusion due to uncertainty over the spatial scales involved. Birds have been shown to alter their flight paths in order to avoid entering an offshore windfarm at distances of up to 6 km (Christensen et al. 2004). As a result, where avoidance rates have been derived from human observations they may represent a substantial under-estimate of total avoidance, as many birds will have taken action to avoid the windfarm before they become visible to observers. The difficulties caused in attempting to draw firm conclusions from such disparate data sources has led to a variety of terms being used to sub-divide avoidance behaviour at different spatial scales.

At a simple level, Cook et al. (2012) and Band (2012) suggest that the total avoidance rate for an offshore windfarm could be considered as (eq. 7):

Total Avoidance =1-(1-Macro × 1-Micro) (eq. 7)

We use this definition as the basis for discussion relating to the different types of avoidance that need to be quantified in order to derive an estimate of total avoidance, and extend it to incorporate meso-avoidance (eq. 8), as defined below.

Total Avoidance =1-(1-Macro ×1-Meso×1-Micro) (eq. 8)

3.2 Defining appropriate spatial scales of avoidance

This section aims to define appropriate spatial scales of avoidance; for detailed review of the evidence for avoidance at these defined scales, see section 5.

A bird may respond to a fixed object, such as a turbine, at any point between the time at which it first observes the object and the time at which it passes or collides with the object, or based on previous experiences of the site. As such, attempts to subdivide avoidance behaviour with reference to spatial scale are largely arbitrary and the different behaviours should be seen as part of a continuum. Nevertheless, such divisions are necessary given the spatial scales over which these behaviours can be recorded. Band (2012) focusses on macro- and micro- avoidance, with a third category, meso-avoidance, fitting in the gap between the two also suggested (Pendlebury, pers. comm.). We consider these scales in turn, with each reflecting an increasing distance between the bird and the turbine blades (Figure 3.1). However, the distances over which these categories of behaviour occur are more difficult to define.

Figure 3.1 Spatial scales over which avian responses to turbines have been recorded

It is also necessary to consider how avoidance rates are applied within the collision risk modelling framework. Expected collision rates (as per eq. 7) are typically derived using estimates of the numbers of birds flying through the windfarm area prior to construction. Therefore, overall avoidance rates need to account for birds no longer entering the windfarm area post-construction ( i.e. birds exhibiting displacement and barrier effects) in addition to avoidance of the turbines themselves. As a result, it is necessary to consider how other effects, such as displacement and barrier effects, may contribute to the overall avoidance rates, as part of macro-avoidance.

We consider how each of these scales may be used to inform collision risk modelling below:

Macro- Band (2012) gives the example of displacement as one impact which may contribute to macro-avoidance. Displacement is typically assessed by comparing numbers of birds in the area of the windfarm to those recorded in a baseline period. However, difficulties in quantifying displacement rates - numbers may vary for many reasons in addition to the development of the windfarm, and it is important that this is considered in an appropriate survey design, for example using a BACI-approach (Masden et al. 2010) - mean that interpreting these data must be undertaken with caution and careful consideration of the survey design (Maclean et al. 2013). Furthermore, published displacement rates can refer to the numbers of birds displaced from the windfarm plus a significant (species-dependent) buffer distance around the windfarm. Consideration must also be given as to whether displacement rates reflect all birds within the windfarm area and buffer, or just those on the water. As collision risk modelling relates only to birds in flight, if displacement rates refer only to birds on the water, they may not reflect macro-avoidance. Relying solely on displacement, as often reported in Environmental Impact Assessments, may therefore underestimate the true scale of macro-avoidance because 1) estimates may not account for birds in flight; and 2) estimates do not account for birds that are displaced from the windfarm area, but remain within the buffer surrounding the windfarm.

In addition to measuring displacement rates, a number of offshore windfarm post construction monitoring studies have used radar to assess the proportion of birds which enter a windfarm area ( e.g. Petterson 2005, Petersen 2006, Krijgsveld et al. 2011). The potential for windfarms to act as a barrier to birds in this way has been widely discussed, mostly in the context of migrants ( e.g. Desholm & Kahlert 2005, Masden et al. 2009), although it may also be of relevance to seabirds commuting between breeding colonies and feeding areas - an area of study that needs addressing with some urgency. Such studies would illustrate changes in flight trajectory amongst birds approaching windfarms and would help to determine the spatial scale over which such responses may occur.

In addition to displacement and the windfarm acting as a barrier, several studies have suggested that some species, notably gulls and cormorants, may be attracted to the area of offshore windfarms ( e.g. Lindeboom et al. 2011, Leopold et al. 2011). The macro-avoidance rate needs to capture the change in bird numbers within the windfarm area resulting from the development of the windfarm site. Consequently, the term 'macro-avoidance', may lead to confusion as, conceptually, the idea of a negative macro-avoidance rate ( i.e. birds being attracted to a windfarm) may be difficult to communicate to stakeholders. For this reason, use of the more neutral term, macro-response, may be preferable as it implicitly covers both attraction and avoidance (Figure 3.2).

Figure 3.2 Range of proportional responses to the presence of an offshore windfarm as they would be incorporated in eq. 2 (above), i.e. a response of -0.1 would reflect an increase in the number of birds present within the windfarm of 10% in comparison to baseline numbers, whilst a response of 0.1 would reflect a decrease of 10% in comparison to baseline numbers, which are sensitive to survey design due to the extent of year on year variation in seabird abundance.

The macro-response of birds to the presence of a windfarm should be defined as the behavioural response taking place outside the windfarm perimeter. It is important that the perimeter of the windfarm is clearly defined. Definitions could be based on characteristics such as turbine rotor diameter, or the inter-array turbine spacing. However, such definitions would vary between sites in relation to the layout and size of turbines used, meaning values for the macro-response rate would be less directly comparable between sites. For this reason, defining the perimeter as extending a fixed distance from the base of the outermost turbines is preferable. The review will define of the perimeter as the boundary of a minimum convex polygon encompassing an area extending from a distance of 500 m from the base of the outermost turbines (see Figure 3.3).

The term macro-response will be used to refer to changes in bird numbers within the windfarm area resulting from the development of the windfarm site, through processes including, but not limited to, attraction, displacement and barrier effects. Where displacement is used to infer a macro-response rate, it is important to be clear whether this reflects displacement from the windfarm only, or displacement from the windfarm plus a buffer. Buffers considered in the assessment of displacement effects typically extend beyond the 500 m around the windfarm perimeter considered here as some birds may respond to the presence of the windfarm at distances greater than this. Measures of displacement that use such buffers may thus underestimate the macro-response rate considered here. As collision risk models refer to birds in flight only, when using displacement rates to estimate a part of macro-avoidance behaviour, it is also important to lend more weight to studies that distinguish the displacement rates of birds in flight and on the water, or those for which it is possible to estimate the number, or proportion, of birds in flight.

Micro- Blew et al. (2008) suggests that micro-avoidance reflects a 'last-second' alteration to a flight path in order to avoid collision with a turbine. Petterson (2005) and Blew et al. (2008) both suggest that birds adjust their flight paths to avoid entering the rotor-swept zone of a turbine and that, therefore, birds may only rarely need to take last second action to avoid collision, possibly as a result of adverse conditions, such as poor visibility. This is borne out by empirical evidence presented in Desholm (2005) and Krijgsveld et al. (2011) (see section 5.3).

Figure 3.3 Schematic illustrating the spatial scales over which micro-avoidance, meso- and macro- responses operate. Dots refer to turbine tower locations (not to scale).

Therefore, it would seem reasonable to define micro-avoidance as a last-second alteration to a bird's flight path in order to avoid collision. For the purposes of observational studies, such last-second avoidance would be expected to occur in a 3-dimensional space within 10 m of the turbine blades ( i.e. at distances of 10 m horizontally or vertically from edges of the turbine blades) - though note that this distance (and consequently the appropriate definition of micro-avoidance) may be refined based on future advances in the techniques used to collect the necessary data (see Figure 3.3). Such behaviour is likely to be recorded relatively rarely.

Meso- Whilst macro-responses reflect behaviour outside the windfarm and micro-avoidance reflects last-second action taken to avoid collision, there is a need to consider a third category, reflecting species responses to turbines within a windfarm (Figure 3.4). Both Desholm & Kahlert (2005) and Krijgsveld et al. (2011) demonstrated that the majority of birds do not pass within 50 m of a turbine. However, some, such as cormorants, may be attracted to structures, which offer potential roosting sites ( e.g. Leopold et al. 2011). For this reason, as in the case of macro-response, it may be more straightforward to talk about a meso-response to turbines than meso-avoidance. The term meso-response should be used to refer to all behavioural responses to the turbines beyond the 10 m buffer around the rotor blades, covered by micro-avoidance, and within the perimeter of the windfarm (see Figure 3.3). This may include, for example the attraction of cormorants to turbine bases as a roosting site, as the base of the turbine would be beyond the 10 m buffer around the rotor blades.

Figure 3.4 Flight trajectories of migrating waterbirds within an offshore windfarm, red dots indicate locations of turbines. Reproduced with permission from Desholm & Kahlert (2005) Avian collision risk at an offshore windfarm. Biology Letters 1: 296-298.

At present, the scale at which data are collected may make it difficult to differentiate between a meso-response and micro-avoidance. Therefore, it is recommended that the term macro-response is used to refer to a response outside the windfarm and within-windfarm response, covering both the meso- and micro-scale, is used to refer to a response occurring inside a windfarm. In response to technological advances, a fuller separation of meso-responses from micro-avoidance is likely to be possible in the near future. For example, it may be possible in future to combine radar monitoring of flight paths through offshore windfarms to capture meso-responses (as in Desholm & Kahlert 2005) with images captured from turbine mounted cameras to capture micro-avoidance (as in Desholm et al. 2006).

3.3 Defining the appropriate 3-D level of avoidance

This section aims to define appropriate 3-D scales of avoidance; for detailed review of the evidence for horizontal and vertical meso-avoidance, see section 5.2.

In addition to occurring over a range of different spatial scales, avoidance behaviour may occur in both the horizontal and vertical planes. Below, we describe how observations of horizontal and vertical avoidance may be collected and the spatial scales which may be relevant to each. This distinction is important given that some methodologies for recording avoidance behaviour, such as radar, may not detect both horizontal and vertical movements, meaning that where only one is recorded, the derived avoidance rate is likely to be an underestimate, which may be offset by an inability to record horizontal and vertical movements occurring concurrently. There is also a need to consider the relationships between avoidance and other effects of offshore windfarms on birds, for example barrier effects and displacement.

Horizontal Avoidance Much of the research into the avoidance behaviour of seabirds in relation to offshore windfarms has focussed on horizontal avoidance, whereby birds alter their flight paths so that they fly around turbines or do not enter the perimeter of the windfarm ( i.e. Desholm & Kahlert 2005, Masden et al. 2009). These data have been collected using a variety of methodologies, notably visual observations ( i.e. Krijgsveld et al. 2011) and radar observations ( i.e. Petersen et al. 2006). We consider that all 3 spatial scales defined here are relevant in the context of horizontal avoidance.

Vertical Avoidance As technologies and survey protocols for monitoring collisions become more developed ( e.g. Desholm et al. 2006, Collier et al. 2011a, 2011b) monitoring of both horizontal and vertical movements around turbines should become more feasible. For radar, however, at greater distance this may be more challenging as detecting both horizontal and vertical avoidance requires the use of both x- and y-band radar. At present, radar monitoring of bird movements in and around offshore windfarms typically focuses on horizontal avoidance behaviour, using horizontal radar ( e.g. Petersen et al. 2006). Where changes in flight height amongst birds entering the windfarm have been estimated ( e.g. Blew et al. 2008) this has been at too coarse a resolution to inform vertical avoidance. However, recent developments in radar technology ( e.g. http://www.robinradar.com/3d-flex/) may make this a more practical solution to investigate vertical avoidance behaviour amongst birds approaching offshore windfarms.

Krijgsveld et al. (2011) demonstrate that a number of species may fly at lower altitudes within-windfarms than outside windfarms and incorporate vertical avoidance behaviour in their estimation of micro-avoidance rates using a combination of visual and radar observations. Their results suggest that a substantial proportion of birds may alter their flight altitudes in order to avoid collision. Given the development of technologies capable of monitoring the movement of birds close to turbines, such as the Thermal Animal Detection System (Desholm et al. 2006), these results suggest that focussing on vertical avoidance at a micro-meso, as opposed to macro, scale may be worthwhile. At a micro-scale, it is likely that vertical avoidance would be captured as part of an evasive manoeuvre.

3.4 Total avoidance rates

In this section, we have produced definitions that are considered to work within the constraints of our current understanding of avoidance behaviour and data collection limitations. It is clear, given the multiple potential components of avoidance behaviour that we have identified (Figure 3.5), that equation 7 is an over-simplification of overall avoidance rates. In future studies it is important to consider how each of these components can be quantified. As technological capabilities advance, the definitions outlined above may become obsolete. However, any refinement to these definitions should be based on the behaviour of the species concerned, rather than artificially induced by methodological constraints, for example, the distance over which observations can be made with the use of binoculars or telescopes.

Figure 3.5 Schematic detailing how different behavioural responses to offshore windfarms may combine to give a total avoidance rate. At each different level birds may respond either vertically or horizontally. Outside a windfarm, both displacement and barrier effects are likely to contribute to the macro-response rate. However, the contribution of displacement to macro-avoidance may be hard to quantify as a result of uncertainty associated with estimating its effects. Avoidance behaviour inside a windfarm is often termed micro-avoidance, however, it may be appropriate to split this term further by considering a meso-response, where birds enter a windfarm but do not pass close to turbines, and micro-avoidance, where birds take last minute action to avoid collisions.

3.5 Recommended Definitions

For the purposes of this review, the definitions we will use for bird behaviour in response to offshore windfarms and turbines are (Figure 3.3):

MACRO-RESPONSE - The response of birds to the presence of the windfarm outside its perimeter, defined as a 500 m buffer surrounding the outermost turbines. Responses may include attraction to the windfarm, displacement from preferred foraging habitat or an alteration to flight paths as a result of seeing the windfarm as a barrier. These may occur in either horizontal or vertical planes, although at present technological limitations mean that it is not possible to measure vertical macro-responses. For this reason, for the purposes of this review, we consider only horizontal macro-responses.

MESO-RESPONSE - A redistribution of birds, or alteration of flightpaths within a windfarm in response to the presence of the turbines. This may encompass both horizontal and vertical responses. These responses are in contrast to micro-avoidance, see below.

MICRO-AVOIDANCE - Last-second action taken by birds flying at rotor height to avoid collision, encompassing both horizontal and vertical movements, within a 10 m buffer surrounding turbine rotor-swept areas.

Due to current methodological difficulties in distinguishing micro-avoidance behaviour from meso-response behaviour, a fourth category is defined for the purposes of this review to act as a proxy for responses to windfarms at these scales:

WITHIN-WINDFARM AVOIDANCE - Encompassing both meso-responses and micro-avoidance to describe how birds within a windfarm respond to the presence of a turbine.

The review focuses on data relating to macro-responses and within-windfarm avoidance. Distinctions between responses at the meso- or micro-scale and horizontal or vertical responses have not been made at this stage as insufficient data are available to support them. Future studies should aim to be able to make such distinctions to improve our understanding of avian avoidance behaviour at offshore windfarms.