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

7. Total Avoidance Rates for Priority Species

In this section, we consider total avoidance rates for each of the five priority species - northern gannet, black-legged kittiwake, lesser black-backed gull, herring gull and great black-backed gull.

7.1 Macro-response rates ( section 5.1)

For gulls, the present evidence base is equivocal, with some studies suggesting evidence for attraction, others evidence for displacement, and others no significant response. Thus, for these species, the balance of evidence suggests a macro-response of 0 ( i.e. no attraction to or avoidance of the windfarm) ( Table 7.1).

Northern gannets typically show a strong macro-response to offshore windfarms. However, differences in survey methodologies make it difficult to arrive at realistic macro-response values by combining data from multiple sources. Based on currently available evidence, we believe that 0.64 to be a reasonable value for the macro-response rate ( Table 7.1). However, it should be noted that this figure is based on data that are most representative of the non-breeding season.

7.2 Micro-response or meso-response rates (sections 5.2 and 5.3)

The review of existing evidence for avoidance rates in relation to offshore windfarms for the key species considered in this study indicated that insufficient data were available to generate separate micro-avoidance or meso-response rates for any of the species of interest.

7.3 Within-windfarm avoidance rates ( section 5.4)

Within-windfarm avoidance rates, representing a combination of meso-responses and micro-avoidance may be derived by comparing observed collisions to those expected in the absence of avoidance (see equation 6 under section 1). Options 1 and 2 of the Band model are mathematically identical (both termed the basic Band model), with the proportion of birds at collision risk height estimated from modelled flight height distributions for option 2 and based on site-specific observational data using option 1. Therefore, it is necessary to use the same avoidance rates for both model options. As the rates derived using option 1 utilise site-specific data, rather than data derived from a generic curve (produced following the methodology of Johnston et al. 2013), we feel that these values are the most appropriate to recommend for use with the basic Band model. With respect to the extended Band model, the rate derived should be acknowledged as, potentially, being precautionary as, at several key sites, it is based on an underestimate of the proportion of birds flying at collision risk height (see Appendix 7). As a consequence, when calculating the avoidance rate by comparing the predicted and observed number of collisions, the resulting value is lower than would otherwise be expected. Therefore, where there is a significant difference between the observed proportion of birds at collision risk height and the proportion predicted to be at collision risk height from modelled distributions, the avoidance rates derived for use with the extended model are not considered appropriate as they will be based on an inaccurate assessment of the number of birds at risk of collision.

An alternative methodology with which to derive a within-windfarm avoidance rate for use with the extended Band model is described by in Annex 1 to this report. Following this methodology, the ratio between the number of collisions expected in the absence of avoidance derived using options 2 and 3 of the Band model is used to modify the avoidance rate derived using option 1 of the Band model. However, this requires knowledge of the flight height distribution ( e.g. to 1m resolution) at the windfarm concerned - as opposed to the proportions of birds assigned to different flight height categories - in order to separate geometric avoidance ( i.e. birds passing the rotor at lower altitudes where the probability of collision is lower) from behavioural avoidance. Whilst it is possible to use this methodology without knowledge of the flight height distribution at the windfarm in question, the result would be that the predicted collision rate using option 3 would be identical to that obtained using option 2. However, this methodology is likely to be of value in the future as data collection techniques improve and detailed flight height distributions are derived on a site-specific basis.

We were able to derive within-windfarm avoidance rates for herring gull and lesser black-backed gull ( Table 7.1). Based on a sample of 526,048 predicted flights through windfarms, we derived an avoidance rate of 0.9959 (± 0.0006 SD) for herring gull based on the basic Band model and 0.9908 (± 0.0012 SD) using the extended Band model. For lesser black-backed gull, the derived avoidance rates were 0.9982 (± 0.0005 SD) and 0.9957 (± 0.0011 SD) respectively, based on a sample of 101,746 predicted flights through windfarms. However, the larger sample size and the fact that data originate from a greater number of sites (see Appendix 7) means that the avoidance rates derived for herring gull are more robust than those derived for lesser black-backed gull. We also derived within-windfarm avoidance rates for large gulls as a group. This group includes all birds positively identified as herring gull (this species accounting for 526,048 of the total of 639,560 flights through windfarms), lesser black-backed gull or great black-backed gull, but also those with uncertain species identification (10,638 predicted flights through windfarms), for example those identified as herring/lesser-black backed gull. For the large gulls group, we derived avoidance rates of 0.9956 (± 0.0004 SD) using the basic Band model and 0.9898 (± 0.0009 SD) using the extended Band model. A comparison of the observed and predicted proportions of birds at collision risk height ( Appendix 7) shows that whilst there are some notable differences in these values, across most sites they are broadly consistent. For this reason, we feel that the avoidance rates derived using both the basic and extended Band models are appropriate to use.

We also derived within windfarm avoidance rates for small gulls (1,589,953 predicted flights through windfarms) based largely on data collected from common gull (746,668 predicted flights through windfarms) and black-headed gull (841,008 predicted flights through windfarms). For species within the small gulls group, we derived within-windfarm avoidance rates of 0.9921 (± 0.0015 SD) for use with the basic Band model and 0.9027 (± 0.0068 SD) for use with the extended Band model ( Table 7.1). However, given significant differences between the proportion of birds observed and predicted to be at collision risk height at a number of key sites, we do not feel that it is appropriate to use the avoidance rate derived for use with the extended Band model for the small gulls grouping. These differences are likely to arise from the fact that the data considered here originate from the terrestrial environment, often close to breeding colonies, whilst the modelled data were collected from the offshore environment.

Finally, we calculated a within-windfarm avoidance rate for all gulls as a group (2,567,124 predicted flights through windfarms). As with the large gull and small gull groups, this incorporated data for individuals with uncertain identification (350,338 predicted flights through windfarms), for example 'gull spp'. For all gulls, we derived an avoidance rate of 0.9893 (± 0.0007 SD) for use with the basic Band model and 0.9672 (± 0.0040 SD) for use with the extended Band model ( Table 7.1). However, as with the small gulls group this includes data for which there were significant differences - due partly to the inclusion of unidentified gulls - between the observed and predicted proportions of birds at collision risk height. For this reason we do not feel that it is appropriate to use the avoidance rate derived for use with the extended Band model for the all gulls groupings.

Insufficient data were available to identify a reliable within-windfarm avoidance rate for northern gannet ( Table 7.1).

It is important to note that where we report the standard deviation around the derived within windfarm avoidance rates, this relates variability between sites and not to uncertainty in the model input parameters. Estimating the contribution of the model input parameters to the uncertainty associated with the derived avoidance rates requires a more detailed understanding of the real range of values associated with each parameter than is available currently.

7.4 Total avoidance rates

Total avoidance rates are also provided in Table 7.1. Ideally, total avoidance rates should be calculated using equation 8 (section 3.1). For gulls, the balance of evidence suggests a macro-response of 0 ( i.e. no consistent attraction to or avoidance of the windfarm). Consequently, the total avoidance rates for these species are equal to the within-windfarm avoidance rates.

As data describing macro-responses to the windfarm are limited, we are unable to estimate the variability around the macro-response rate. For this reason, whilst we are able to provide an estimate of variability around the within windfarm avoidance rates, we are unable to provide an estimate of variability of uncertainty around the total windfarm rates.

Table 7.1 Derived avoidance rates for priority species and current knowledge gaps based on the review of available data. Empty cells indicate a lack of robust and/or consistent data on which to base conclusions. Colour coding indicates confidence in presented values (based on sample size, representativity of data): green = highest, orange = intermediate, red = lowest ( i.e. not suitable for use in CRM). Confidence in total avoidance rates reflects the lower of the confidence ratings given for macro-responses and within-windfarm avoidance rates.

7.5 Recommended avoidance rates

Please note that these recommendations apply to the five priority species only - northern gannet, black-legged kittiwake, lesser black-backed gull, herring gull and great black-backed gull - they are not intended to be applied to seabirds more generally.

Whilst we have estimated within-windfarm avoidance rates to four decimal places, current guidance from SNH is that expressing avoidance rates to more than three decimal places is unwarranted ( SNH 2013). Given the inherent uncertainty in the data we feel that this is a sensible approach to apply to total avoidance rates. For this reason, we round within-windfarm avoidance rates down to three decimal places when deriving recommended total avoidance rates.

• A macro-response rate of 0.64 is recommended for northern gannet ( section 5.4). However, no data were available to derive a within-windfarm avoidance rate for this species ( section 5.3). Given that there is consistent evidence for high macro-avoidance, and considering the at-sea ecology of northern gannet and gulls (section 6.3.5), we feel that there is no reason to suppose that the total avoidance rates for northern gannet should be less than those for all gulls (as opposed to large gulls). A total avoidance rate of 0.989 is thus recommended for use with the basic Band (2012) collision risk model. This would reflect a within windfarm avoidance rate of 0.9703. We acknowledge that this is precautionary, but in the absence of more species-specific data, we feel it is appropriate. However, given the evidence available at present, we are unable to recommend an avoidance rate for use with the extended Band (2012) collision risk model.
• No consistent evidence of macro-avoidance was found for black-legged kittiwake ( section 5.4). It was not possible to derive species-specific within-windfarm avoidance rates for black-legged kittiwake ( section 5.3) . However, as black-legged kittiwake have similar wing morphologies (wingspan, wing:body aspect ratio, wing area: Robinson 2005, Alerstam et al. 2007), flight speeds (Alerstam et al. 2007) and flight altitudes (Cook et al. 2012, Johnston et al. 2014b) to black-headed and common gulls, which contribute the majority of records for the small gulls group, the within-windfarm avoidance rates derived for the small gulls group were considered appropriate for this species. A total avoidance rate of 0.992 is thus recommended for the basic Band model. However, given the evidence available at present, we are unable to recommend an avoidance rate for use with the extended Band (2012) collision risk model ( section 5.3).
• No consistent evidence of macro-avoidance was found for lesser black-backed gull ( section 5.4). Whilst it was possible to derive species-specific within-windfarm avoidance rates for lesser black-backed gull, these were based on limited data and thus the within-windfarm avoidance rates for large gulls were considered more appropriate for use for this species ( section 5.3). A total avoidance rate of 0.995 is thus recommended for use with the basic Band model and a total avoidance rate of 0.989 for use with the extended Band model.
• No consistent evidence of macro-avoidance was found for herring gull ( section 5.4) and thus total avoidance rates reflect species-specific within-windfarm avoidance rates. A species-specific total avoidance rate of 0.995 is thus recommended for use with the basic Band model and a total avoidance rate of 0.990 for use with the extended Band model ( section 5.3).
• No consistent evidence of macro-avoidance for great black-backed gull ( section 5.4). It was not possible to derive species-specific within-windfarm avoidance rates for great black-backed gull. Given the taxonomic similarity between species within the large gulls group, the avoidance rates derived for use with this group were considered to be appropriate for great black-backed gull ( section 5.3). A total avoidance rate of 0.995 is thus recommended for the basic Band model and a total avoidance rate of 0.989 for use with the extended Band model.

At present, the evidence available does not enable us to recommend a robust avoidance rate for northern gannet or black-legged kittiwake for use with Band model option 3. This does not imply that option 3 is not suitable for these species, and given the programmes of work currently underway in the offshore environment, it is envisaged that an appropriate rate will be derived in the near future. Note, while it is not possible to recommend a robust avoidance rate for use for these species at this time, this does not preclude presenting a no-avoidance collision estimate using option 3 alongside collision estimates derived using option 1 and/or option 2 (with or without using the avoidance rates recommended here) to inform on likely collision risk.

Table 7.2 Recommended total avoidance rates for use in the basic and extended Band models with each of the five priority species.