Scottish Marine and Freshwater Science Volume 5 Number 12: Strategic assessment of collision risk of Scottish offshore wind farms to migrating birds

5. Discussion

5.1. As with any collision risk modelling analysis assumptions regarding bird flight patterns and behaviour have been made. With the focus on assessing migration mortality a key assumption relates to the extent of migration routes for each species. This has been approached separately for seabird and non-seabird species. The starting point for defining the migration corridors of non-seabird species was the figures provided in the SOSS-05 report. These have been refined where possible and also modified to reflect passage through Scottish waters. For most species very little is known about the routes followed, and this is reflected in the wide corridors used. The effect on the collision mortality estimates obtained using these wide corridors may either be to generate estimates which are too high or too low: the former if 'real' migration routes do not traverse any Scottish offshore wind farms, the latter if the 'real' routes are concentrated through areas which contain wind turbines. The wide corridors used here increase the predicted number of wind farms which will be encountered, but also reduces the proportion of the passage population calculated to be at risk. The relationship between collision risk and migration front width is linear; a doubling or halving of the migration front leads to a halving or doubling of the collision risk respectively if wind farms are encountered (assuming a uniform distribution of birds across the full corridor width).

5.2. For seabirds a slightly different approach has been taken. Five coastal corridor widths were defined and each seabird species assigned to one. Furthermore, three alternative suites of corridors were defined which differed in how closely the inner edge followed the coastline across inlets. The passage population of each species was then divided into components which pass either the east or west coasts. These proportions were based on knowledge of where birds are heading and coastal observations where available.

5.3. Obtaining reliable and up to date population estimates is an essential requirement for impact assessments. This is typically a challenge even for resident species with limited distributions. For migratory species with ranges spanning 1,000s of kilometres and several countries, the challenge is considerably greater. In this assessment the most recent and reliable estimates have been used, however these estimates are not necessarily the only ones available. However, should population estimates be updated in the future the impacts as presented here can still provide a guide, since as long as the migration corridor assumptions remain valid, the collision estimates will scale with the revised population size ( e.g. if the population doubles the collision mortality will double, and vice versa.)

5.4. Distribution of seabirds across the coastal corridors was modelled as either uniform ( i.e. an even distribution across the complete corridor width) or using a skewed distribution (negative binomial) which increased the proportion predicted to fly near the coast. Each combination of coastal corridor distance from shore (near/mid/far) and width distribution (uniform/negative binomial) generated different collision estimates due to the variation in the extent of overlap between each corridor and the offshore wind farms. For example, herring gulls were predicted to be at greater risk of collisions when modelled as being uniformly distributed across migration corridors located near or mid shore. However, the highest collision estimates for each suite of coastal bands was obtained when using the skewed negative binomial width distribution ( Figure 60). Similar patterns were seen for Arctic skua, little tern, cormorant, common gull, common tern, great black-backed gull, lesser black-backed gull, red-throated diver and shag. Conversely, black-headed gull and kittiwake were predicted to be at decreasing risk of collision as the coastal strip moved farther offshore (Figures 51 and 61). These patterns appear to reflect the width of the migration corridor to which each species has been assigned. The first group were all assigned to the 0 - 10km or 0 - 20km coastal bands. When combined with the near-shore option these result in lower numbers of individuals passing through the wind farms. When the coastal bands are moved farther offshore, the number predicted to pass through the offshore wind farms increases leading to elevated collision risks. Black-headed gulls and kittiwakes were both assigned to the 0 - 60km coastal strip. The proportional overlap of this strip with the Scottish offshore wind farms falls as the coastal strip moves farther offshore, thereby lowering collision risk. It is not possible to determine which of the alternative scenarios provides the closest representation of migration activity for any given species. However, this analysis has highlighted that improved understanding of the factors considered here is likely to help refine future estimates of collision risk and guide identification of wind farm sites on the basis of risks to migrating birds.

5.5. The recent addition of flight height distribution data to the calculation of the probability of collision (Band 2012; Option 3) was used for this assessment alongside the previous method (Option 1) and an intermediate one (Option 2) which makes partial use of the additional flight height data. Taking account of the magnitude of altitudinal overlap between turbine rotor blades and flying seabirds reduces the collision mortality estimates for all species with one unlikely exception: razorbill. One note of caution is required however, since the flight height data may not be representative of seabird migratory flight, since in most instances it will contain observations made at other times of the year when birds may be flying at different heights.

5.6. The higher collision rate estimated for razorbill appears to be due to the fact that the overall estimate of birds at potential collision height, as used in CRM Option 1, is very low at only 0.4%. Indeed, the flight height distribution data (Cook et al. 2012) reveal that over 99% of individuals fly at less than 10m. However, rather than the modelled flight distribution decreasing monotonically from there ( i.e. with no 'wiggles'), there are some very small increases in the modelled proportions at heights which overlap with turbine rotors. When these data are used in the collision model these generate the slightly higher collision probabilities seen in the Option 3 mortality estimates. Given the flying behaviour of this species, this result is considered to be an artefact (possibly of flying observations made close to shore or in relation to survey vessels) and not reliable.

5.7. For all other seabird species for which flight data were available, the tendency for birds to fly close to the sea surface with numbers decreasing with increasing altitude, results in considerably lower collision estimates when used in the CRM. The exact magnitude of reduction varies between species, from a 26% reduction for black-headed gulls to over 90% for shag, kittiwake, common tern and Arctic skua. Uncertainty in flight height observations is provided through the modelled confidence range. In most cases, even using the upper confidence estimate the overall reduction in collision estimate which comes with CRM Option 3 is such that this upper bound is still less than that obtained using Option 1.

5.8. Many fewer simulations were conducted for the non-seabird species, reflecting the generally poorer level of understanding of migration behaviour in these species. There is also no large dataset of flight height data currently available with which to undertake an Option 3 approach for estimating the probability of collision. Thus, only CRM Option 1 was available for this assessment.

5.9. While the current assessment provides an overview of potential collision risk for non-seabird species, it is very sensitive to the assumptions made regarding the migration corridors used and the offshore wind farms predicted to be encountered. Many of the species considered travel across large tracts of sea before making landfall in the UK. Crucial to calculating the potential for collision is knowledge of how wide the land fall destination range is. However, this information is not available and thus it was necessary to assume wide migration corridors with birds spread evenly across them. The proportion of these migration corridors which overlap with Scottish offshore wind farms is generally small. Hence the proportion of each passage population calculated to be at risk of collision was also small. This is reflected in the overall low collision mortality estimates obtained ( Table 18). Furthermore, because of the small overlaps the estimates are fairly insensitive to the exact wind farm dimensions used, so variations in final layouts compared with those currently planned are unlikely to affect the results to any appreciable extent.

5.10. Assuming an indicative threshold value of 1% of the passage population, no non-seabird species had collision mortality estimates (at 98% avoidance) that were of concern