Offshore wind - birds on migration in Scottish waters: strategic review

Discussion

Through this review, we have set out to assess the migratory behaviours of designated features of SPAs in the UK and consider the implications of these behaviours in relation to potential collision risk with offshore wind farms. Drawing from this review, we have summarised suitable input parameters for each species (Electronic Appendix 5) to feed into the migrant collision risk modelling tool being developed as part of work package 2. However, despite considerable advances since previous reviews of this topic (Wright et al. 2012; WWT Consulting, 2014), particularly in relation to tracking of birds on migration, significant knowledge gaps remain, and there is still uncertainty surrounding values for some parameters, for many species. However, the data presented here, and summarised in Table 6 reflect the best available evidence for collision risk modelling of migratory species in relation to offshore wind farms.

Confidence levels vary markedly between parameters (Table 6), with greatest confidence in the estimates of flight speed, and lowest confidence in estimates for the avoidance rate. This reflects the weight and quality of evidence for each parameter. For most species, we were able to obtain estimates of flight speed from a variety of sources (e.g. GPS, Radar, Laser Rangefinder), and the estimated speeds were generally consistent across studies. Consequently, we often had a high degree of confidence in the recommended values. Whilst estimates of species flight heights on migration, over the sea, are available for many species, these data often came from fewer studies, often with limited sample sizes. Furthermore, estimates for species flight heights were typically reported as a mean or median value, with some measure or uncertainty. This makes it less straightforward to assess the likely proportions of birds at collision risk height than would be the case if a continuous flight height distribution were available. Consequently, we have less confidence in the estimates for this parameter than is the case for estimates of flight speed. Finally, in most cases, estimates of avoidance behaviour are generally derived from onshore studies, and it is unclear the extent to which these are transferable to the marine environment. Whilst they are likely to reflect the best available evidence at present, this uncertainty means that we generally have very low confidence in the recommended value, except in situations where there are alternative sources of evidence suggesting a strong avoidance response, as is the case with divers (Mendel et al. 2019) and seaducks (Desholm & Kahlert, 2005).

The sensitivity of collision risk models to their input parameters has been widely acknowledged (Chamberlain et al. 2006; Masden & Cook, 2016; Masden et al. 2021). In considering the importance of any knowledge gaps and uncertainties, it is important to consider the relative sensitivity of the model to these parameters. Recent analysis has highlighted four key parameters to which the model is likely to be sensitive – the total number of birds passing through a wind farm, the speed at which these birds are flying, the height at which they are flying and the extent of any avoidance behaviour (Masden et al. 2021).

Number of birds passing through a wind farm

The total number of birds passing through a wind farm is a function of the size of the population of the species concerned and the extent and location of the corridor through which they migrate. There may be considerable uncertainty about both of these factors, though this depends on the species concerned.

Defining migration corridors for species can be challenging. For many, it is unclear whether birds migrate across a broad front, or within a narrow corridor. Understanding this will have important implications for determining the number of birds at risk of collision with offshore wind farms. If birds are assumed to migrate across a broad front, but in reality, migrate within a narrow corridor, then the risk posed by a wind farm on the migration route will be under-estimated, but the risk posed by wind farms elsewhere will be over-estimated. The reverse is true in the case of birds which are assumed to migrate within a narrow corridor, but instead migrate across a broad front.

For the purposes of this report, we follow the approach of Wright et al. (2012) in defining migration corridors based on ring recoveries of the species concerned. A consequence of this is that migration corridors are assumed to be fairly broad. However, the rapid expansion of GPS tracking means that we are now able to consider how valid an assumption this is likely to be. For some species, like Whooper Swan (Griffin et al. 2011), which winter in a small number of well-known sites in the UK, it is possible to track a relatively representative sample of the population. Comparison of GPS tracks and the migration corridors defined using ringing data suggests that while birds may migrate within a narrower band than is assumed by this approach, reliance on ringing data would not lead to a gross over-estimate of the migration corridor for this species (Appendix 1).

For species with widespread distributions, gaining a representative sample of the birds migrating to, or through, the UK is challenging. However, GPS data from these species can still yield useful insights into the likely extent of migration corridors in these species. For example, GPS tracks from relatively small samples of oystercatchers and shelducks wintering on the East coast of England show substantial individual variation in the routes taken (Clewley et al. 2021; Green et al. 2021; Appendix 1). Such data would imply that, for these species, the assumption of a broad front migration, as defined by ringing data, is likely to be valid. Indeed, for the 15 species and populations for which we were able to obtain GPS tracking data, the migration corridors defined using ringing data appear to be a realistic representation of the routes taken by these species (seven goose species or subspecies, Bewick's Swan, Whooper Swan, Shelduck, Oystercatcher, Ringed Plover, Osprey, Hen Harrier and Short-eared Owl). However, as is the case with geese and swans, there is the potential to better define these routes, and therefore reduce uncertainty about the proportions of birds passing through any given wind farm, through the use of data obtained using approaches such as GPS tracking.

Estimates of the size of the population of these species are available from a variety of sources (Wetlands International, 2021; Forrester et al. 2007; Humphreys et al. 2021; Woodward et al. 2020). However, the quality of the data underpinning these estimates can be highly variable. Whilst some are based on regular censuses (e.g. Brides et al. 2021), others are based on extrapolations from historic count data, and may be considerably less robust (e.g. Green et al. 2019). Furthermore, whilst estimates are available for the number of birds breeding or wintering in the UK (Woodward et al. 2020), for some species, especially waders, substantial numbers may pass through the UK during migration between breeding and wintering grounds (Wernham et al. 2002). These species may spend substantial portions of the year in areas where ecological monitoring is extremely limited (Proença et al. 2017), leading to significant uncertainty about the population size of the species concerned.

These factors mean that there is likely to be considerable variation between species in the uncertainty surrounding estimates of the total numbers of birds passing through any given wind farm.

Flight speed of migrating birds

The Band Collision Risk Model for seabirds (Band, 2012) uses flight speed to estimate the total number of birds likely to pass through a wind farm, and the probability of any bird which passes through a turbine rotor sweep colliding with one of the blades. In contrast, the model for migrants only uses flight speed to estimate the probability of a bird passing through the rotor sweep, colliding. Consequently, it is likely to be less sensitive to flight speed than the seabird model (Masden et al. 2021). However, it is important to ensure robust estimates of flight speed are available.

Estimates of species' flight speeds are available from a variety of sources including radar (Bruderer & Boldt, 2001), ornithodolite (Pennycuick, 1997) and GPS tags (Mellone et al. 2012). Where possible, the speeds reported in this review reflect data collected during migration from the marine environment. It was possible to obtain published estimates of speed for all but seven of the species covered in the review (Purple Sandpiper Calidris maritima, Dotterel Charadrius morinellus, Avocet Recurvirostra avosetta, Stone-curlew Burhinus oedicnemus, Slavonian Grebe, Great-crested Grebe and Corncrake Crex crex). For these species, it was necessary to make recommendations about flight speed based on data from related species.

For many species, flight speed estimates were obtained from a single study, often with limited sample size, resulting in low confidence in the reported values. Where data were obtained from multiple studies and/or using multiple approaches, these were broadly consistent with each other. However, there was evidence of potential differences in speed during the spring and autumn migrations, potentially reflective of either prevailing weather conditions, or the need for birds to return to breeding grounds early in order to secure optimal territories. There is a need to consider whether, and how, such differences should be accounted for when assessing migrant collision risk.

Flight height of migrating birds

Estimating the proportion of birds at collision risk height has long been seen as a crucial part of estimating the potential risk of collision (Krijgsveld et al. 2009). A variety of methodologies have been developed to enable the estimation of species' flight heights (Largey et al. 2021). For seabirds, it has become standard practice to summarise data from surveys and GPS tracking as continuous flight height distributions (Cleasby et al. 2015; Johnston et al. 2014; Ross-Smith et al. 2016) which can be assessed relative to the known height of offshore turbines in order to quantify the proportion of birds at collision risk height. However, this is not the case for migrating birds, for which data are often reported as a mean/median with an associated standard error or confidence interval. It cannot be assumed that these data are normally distributed, making assessing the proportion of birds at any given height extremely challenging, especially given that many of these values are within, or close to the range of heights expected for modern turbines.

Estimates of species' flight heights were available for fewer species than was the case for flight speeds. There are two key reasons for this. Firstly, it was important to ensure that the data incorporated in this review usually were restricted to migratory movements over the sea. Altitudes reported for birds moving over the terrestrial environment often do not account for underlying topography, meaning that in these cases data would not be reflective of the heights reached by birds in relation to wind turbines. This means that whilst it was possible to consider flight speeds from the terrestrial environment in the absence of data from the marine environment, this is generally not the case in relation to flight heights. Secondly, given the challenge of collecting flight height data in the marine environment, such studies often rely on the availability of GPS data. Battery life is often an important consideration when collecting data as part of GPS studies. Given the additional power requirements necessary to obtain flight height estimates from GPS, these data have often been restricted to larger species such as raptors, geese and swans (Griffin et al. 2016; Mellone et al. 2012). As technology develops and tag size decreases, it is becoming possible to collect GPS flight height estimates from smaller-bodied species (Clewley et al. 2021; Green et al. 2021). These data can be analysed using approaches similar to that set out in (Ross-Smith et al. 2016) enabling a clearer assessment of the proportion of birds at collision risk height in relation to different turbine designs. However, existing data suggest that there may be spatial variation in the proportions of birds at collision risk height (Griffin et al. 2011), which could be accounted for in the analysis of these data, reducing uncertainty in this parameter further.

Avoidance rates and behaviour

The avoidance rate is often seen as the key parameter in relation to collision risk models (Chamberlain et al. 2006). Consequently, there has been significant effort devoted to quantifying avoidance, both through the review and analysis of existing data and, the collection of new data (Skov et al. 2018).

Birds are generally considered to respond to wind farms and turbines in relation to three spatial scales – macro (response to the wind farm), meso (response to individual turbines) and micro (last-second action to avoid collisions) (Cook et al. 2014; May, 2015). Through the analysis of post-construction monitoring data (Dierschke et al. 2016), the spatial modelling of bird movement data from GPS (Schaub et al. 2020; Thaxter et al. 2018) and observations with camera systems (Desholm et al. 2006; Skov et al. 2018) it is possible to quantify the proportion of birds responding to the wind farm itself, and individual turbines within the wind farm. However, the avoidance rate used by collision risk models must account for both the behaviour of the birds, and error in how the model is structured and parameterised (Masden et al. 2021). Consequently, whilst we have strong evidence that several species, including seaducks and geese, show strong avoidance of offshore wind farms when migrating (Desholm & Kahlert, 2005; Masden et al. 2009; Plonczkier & Simms, 2012), these data on their own do not reflect the avoidance rate as used by collision risk models. Instead, it is often necessary to rely on data collected from the onshore environment where it is possible to compare observed collision rates to those that would be expected in the absence of collision (Cook et al. 2018, 2014). As it is unclear how transferable these rates may be between the onshore and offshore environments, and how representative the data are of the full range of conditions experienced by the species concerned, confidence in the reported rates is typically very low, except where supported by evidence of strong macro avoidance from tracking or radar studies. As technology develops, and deployment of devices such as GPS and camera systems becomes more widespread, there is a need to consider how the resulting data can be better integrated into the assessment of collision risk.

Effects of climate change on migration

Offshore wind farms are expected to have a typical lifespan in the region of 25 years. Over this time period, there is the potential for climate change to influence species' migratory behaviours, not least in relation to the routes which they historically use (Robinson et al. 2009). Consequently, it is important to consider the process through which climate change may influence species migration routes, and the potential implications for interaction with offshore wind farms and risk of collision with turbines.

Taking a wide overview, climate change is generally simplified as alterations to normal weather and temperature patterns (IUCN 2012), shifting these away from the overall conditions that species expect. Such changes in climate are already known to have had an impact on birds (Crick 2004). For species that regularly make decisions on when to move and how to navigate based on climatic variables (Able, 1973), such as melting ice signalling the right time to move to breeding grounds (Lameris et al. 2018), or those that have to land when hurricanes form in the middle of their flight paths (Huang et al. 2017), the risk of disturbance to these patterns is more likely with climate change. With flight paths and migration routes changing, even if only marginally, there will be more occasions that birds may be at increased risk of collision, particularly for species that move through the North Sea due to the large numbers of areas highlighted for offshore wind development^[15].

Impacts on migration from changes to wind patterns and increasing incidence of storms

That meteorology must be "favourable" for bird migration has long been accepted and studied (Smith, 1917) and with increasing shifts from what we call 'normal' due to climate change, subsequent impacts and detours are not unexpected (Shamoun-Baranes et al. 2017). Changes to wind patterns is our first starting point to consider how migration routes through waters around the UK might alter (Liechti, 2006), as is the increasing incidence of severe storms (Butler, 2000; Newton, 2007). Under these circumstances, birds are likely to move from their expected, and usually well known, flight paths, and may find themselves moving to more coastal waters or further into the marine environment – thus could increase their exposure to collision risk.

Some migratory birds wait for optimum conditions where "wind assistance" can be utilised to move themselves across areas such as the North Sea and when facing uncertain wind conditions they may take unexpected and unknown routes, particularly at night (Bradarić et al. 2020), which might reduce their capacity for visual avoidance of turbines. Given birds often time their migration for the periods of best tail-winds (Alerstam, 1990), misunderstanding new air flows and meteorological conditions could also have implications for collision risk.

Shifts in migratory altitude is another consideration for when birds run into the possibility of coming in range of turbine rotors rather than flying outside the blades' airspace (Bowgen & Cook, 2018; Thaxter et al. 2018). Whilst birds do tend to migrate lower over the sea than over land, bad weather will force individuals to move to lower altitudes and even to settle on the water increasing their potential exposure to collision risk (Drewitt & Langston, 2006).

Impacts on migration from changes to air and sea temperatures

Another impact from climate change on migratory movements can be seen through changes to the ambient (air) and sea surface temperatures. Fluctuations in ambient temperature can shift the phenology of birds' life histories (Gordo, 2007). Most commonly these are through either better breeding/winter habitat conditions at different latitudes resulting in different migration routes or shifts in birds' timings to migrate earlier or later in the year causing them to face different meteorology (Gordo 2007). Often termed "short-stopping", current research indicates that climate change is expected to cause shifts in the distribution of birds wintering in Europe towards the northeast (Elmberg et al. 2000) which from the UK would move many birds across wider expanses of the North Sea. This phenomenon is exemplified by Bewick's Swan Cygnus columbianus bewickii (see below), which not only has shifted its wintering range eastwards, but also demonstrates 'short-staying', where the amount of time spent on the wintering grounds has reduced over time (Nuitjen et al. 2020). These responses to warming temperatures can be a product of either generational change over time (in the case of short-staying), or both generational shifts and individual plasticity (Nuitjen et al. 2020). This suggests that while in some cases, responses to climate change in terms of changing migratory patterns can be relatively quick (individual plasticity), in other cases migratory patterns may be expected to change more slowly (generational change).

Changes to sea surface temperature are likely to mean that marine food resources may not be present in the same locations as birds expect when needing to refuel during migration (Fujii, 2012). Such changes in water (or subsequently intertidal sediment) temperature are known to affect the optimum habitats for various fish (Mitchell et al. 2020) and marine invertebrates (Beukema et al. 2009; Kendall et al. 2004) preyed upon by birds. With generally warming temperatures, prey moves away from intertidal and surface areas into deeper and cooler waters (Hiddink et al. 2015). It should also be noted that associated sea-level rise from increasing temperatures (i.e., polar ice melt) will also change where invertebrate food resources may be found by birds utilising stopover sites along migration corridors (Fujii, 2012).

Species which normally use stopover sites along the UK coastlines or follow migration paths that coincide with food resources (Howard et al. 2018) could find themselves needing to move to new areas which may either have wind turbines or to pass through such areas. Wind turbines are often further offshore where waters are cooler and thus provide suitable habitat for such moving food resources (Wright et al. 2020). It should be noted that studies have shown that species that forage in the water column (e.g., auks, shearwaters, shags, cormorants, and gannets) are less likely to be as affected by shifting prey depths than species which feed in the surface layers (e.g., terns, gulls, skuas, storm-petrels) (Mitchell et al. 2020).

Impacts on specific species

No information on direct impacts of climate change and wind farm collisions are available in the current literature and so these example case studies indicate where migration behaviours of species that interact with the marine environment have changed as a result of weather or climate change.

Storm driven route changes – A study of Whimbrels migrating down through the Americas investigated the overlap of migration routes and storm activity. It showed that over half of the storms encountered resulted in grounded birds on islands and several routes around core storm activity were taken regardless of length (Watts et al. 2021).

Short-stopping – Bewick's swans have been identified as "short-stopping" on migration to their wintering grounds and "short-staying" at their wintering sites during 1970–2017 and 1989–2017 respectively (Nuijten, et al. 2020). This is thought to be due to climate change promoting more favourable air temperatures across their range, in particular with the 5.5⁰C isotherm having shifted eastwards over time. Bewick's swans may have adjusted their migration routes to pass in a narrow corridor across the North Sea to concentrate on more southerly sites in the UK and thus may find themselves in conflict with wind farm developments in the southeast, but less so in other parts of the UK.

Temperature-related shifts in migration timing and sites – A review of climate change effects on European duck populations highlighted advancing spring migration for many duck species likely resulting from warmer winter conditions promoting better body reserves and thus earlier departures from wintering grounds (Guillemain et al. 2013). The same review also highlighted general northward and eastward shifts in geographic ranges due to better access to ice-free water. Tringa sandpipers (Greenshank, Spotted Redshank and Wood Sandpiper) have also been identified as shifting to earlier spring migrations after warmer winters and whilst they had all delayed autumn migration, this was not significantly linked to climate by the study (Anthes, 2010).

Summary of potential climate change effects on migration

Climate change has the potential to cause route alterations and alter migration timing of migratory species moving through UK waters. Some species will be able to adapt to changing routes and prey patterns more easily and/or more quickly than others, in particular those with generalist diets or those which face fewer life history constraints during their annual cycle. Collision risk models need to be suitably flexible to adapt parameters to changing migratory patterns in light of climate change and should be reviewed regularly to identify if parameters are likely to have shifted as a result of climate change.

Suggestions for future research

Additional GPS tracking studies

For all but the larger-bodied wildfowl and raptors, ringing data provides the only current information on migration corridors. High resolution tracking of larger species suggests that different species have different widths of migratory corridors – some migrate through a narrower corridor (e.g., Greenland White-fronted Goose, Taiga Bean Goose), while others migrate on a broad front (e.g., Oystercatcher). However, for the majority of species, especially waders and ducks, there are still extensive knowledge gaps as to migration corridors. Recent developments in tracking smaller-bodied species such as the large waders (Oystercatcher) and Shelduck have begun to fill these knowledge gaps, suggesting that relatively broad-front migrations, even from the same population, are not uncommon. Lighter-weight GPS devices and the ability to make shorter-term tag deployments using glue-mounted tags or short-term harnesses has expanded the capacity to obtain high resolution oversea crossing data from species as small as Redshank.

Future research should aim to expand high resolution tracking on both autumn and spring migrations for not only the large-bodied waders and ducks (and also some goose species for which there are still knowledge gaps yet the tracking technology and safe, long-term, deployment techniques already exist), but also medium-sized waders. These data will also provide much needed information not only on migration corridors, but also on which species are diurnal vs nocturnal migrants oversea, on flight speed, and for certain devices, will also provide information on flight heights.

Further analysis of existing GPS tracking data

The availability of high-resolution GPS data for a range of species offers the potential for more detailed analysis of migratory movements that could, ultimately, reduce uncertainty surrounding potential collision risk. A key priority in relation to this is reducing uncertainty surrounding the proportion of birds at collision risk height. At present, these estimates are largely based on precautionary assessments informed by mean or median flight heights derived from GPS data where available. However, analysis of GPS flight height estimates, following approaches such as those set out in Ross-Smith et al. (2016) or Cleasby et al. (2015) would enable us to better understand the likely proportions of different species at collision risk height, and how this may vary in relation to the location of the wind farm concerned. Similar analysis could be undertaken in relation to flight speed, with a particular focus on seasonal differences in flight speed, and understanding the drivers for these differences (e.g., due to prevailing weather conditions or, the need to reach breeding grounds early).

In addition to improving knowledge of how birds migrate, there is the potential to use existing GPS tracking data to elucidate when they migrate, and how this may influence collision risk. For example, seasonal phenology of oversea crossings may vary year to year depending on conditions encountered at wintering, passage, or breeding sites (Amélineau et al. 2021). Furthermore, there is concern that birds which migrate at night may be attracted by the lighting on offshore wind farms (Hüppop et al. 2006), and therefore be at greater risk of collision. Analysis of GPS data could highlight areas where there is a particularly high risk of birds moving at night, where they may be attracted by turbine lighting. Such information could be incorporated into assessments of collision risk through revision of avoidance rates.

At present, such analyses could be undertaken for species for which high resolution GPS data are available - Bewick's Swan, Whooper Swan (continuous, 3D GPS data at one second intervals), Taiga Bean Goose, Pink-footed Goose, Greenland White-fronted Goose (continuous 3D GPS data), Icelandic Greylag Goose, Svalbard Barnacle Goose (continuous 3D GPS data), Greenland Barnacle Goose (continuous 3D GPS data), Shelduck, Oystercatcher and Curlew. However, as technology develops, and data become available for a greater range of species, there may be the potential to expand this further.

Model development

As technology develops, it will be important to ensure that any migrant collision risk modelling tool also develops in order to take account of these changes. Initially, this may involve incorporating different flight speeds for spring and autumn migrations, where data allow it. However, as we gain a better understanding of flight heights this may include incorporating continuous flight height distributions in a similar manner to the existing extended Band (2012) model, but may eventually account for spatial patterns in species' flight heights over their migration route.