Scottish Marine and Freshwater Science Volume 6 Number 10: At-Sea Turnover of Breeding Seabirds - Final Report to Marine Scotland Science

1. Introduction

The Scottish government has set a target of 100% of Scottish demand for electricity to be met by renewable sources by 2020 and an interim target of 50% by 2015. Offshore wind will be a key contributor to the renewable portfolio, and a Marine Plan identifies areas of development in the short term (up to 2020) and medium term (beyond 2020; Marine Scotland 2011; Scottish Government 2013). Some of these areas host important populations of seabirds that are protected by the EU Birds Directive. Offshore renewable developments have the potential to impact on protected seabird populations, notably from collisions with turbine blades and through displacement from important habitat.

In undertaking assessments of potential impacts of offshore wind farms on seabirds, interest lies in estimating the number of birds that will be present in a particular area of sea ( e.g., the footprint of a proposed offshore wind farm) at a particular time, relative to the total number of birds that will use that area of sea at any point during the breeding season. This relative use of an area at a given time in relation to the rest of the breeding season is termed 'turnover', relating to the turnover of individual birds using a particular area over time. Estimating turnover is important because estimates of the number of birds that may be affected by offshore renewable energy developments typically involve a limited series of at-sea surveys of fixed areas (potential wind farm footprints). These surveys effectively provide a snapshot estimate of the number of birds using that area at different times during the breeding season. Therefore, there is a need for better understanding of the extent to which these snapshot estimates underestimate the total number of birds using the area over the entire breeding season. This project, (1) reviews input parameters required to estimate turnover for remaining key species to establish data gaps; and (2) estimates turnover within selected areas (wind farm footprints) within the Forth/Tay offshore wind farm development area for selected species with sufficient data. To do so, we build upon previous work conducted in a larger project estimating the effect of displacement on breeding birds in this region (Searle et al. 2014).

In this project we consider how the turnover of birds varies by species (black-legged kittiwake Rissa tridactyla; common guillemot Uria aalge; razorbill Alca torda; Atlantic puffin Fratercula arctica), and with other biological and methodological parameters. Inevitably, the degree of turnover of individuals using an area over the breeding season will be influenced by the extent to which birds tend to return to the same foraging location repeatedly through time (termed 'site fidelity'), and the spatial scale over which fidelity to particular foraging locations operates (termed 'spatial scale of site fidelity'). We explore how estimates of turnover are influenced by the level of site fidelity - with individuals displaying behaviours along a scale ranging from no site fidelity (foraging locations are selected independently on each day) to complete site fidelity (the same foraging location is used throughout the breeding season). This necessarily raises the question as to how we define the appropriate spatial scale over which site fidelity operates. The spatial scale of site fidelity will vary by species, and is likely to vary seasonally in response to environmental conditions. However, empirical data on the spatial scale of site fidelity for foraging birds is scarce, so in this project we consider a range of scales over which it is assumed fidelity to foraging sites operates in each species.

We also consider how the scale and location of survey effort affects estimates of turnover. To do so, we vary both the location and size of potential wind farm footprints over which observations of individuals occur. In addition, when at sea, boat or aerial surveys are conducted, observed birds are classified as far as possible according to behaviour - either as resting on the sea surface, foraging, or flying over the area. These distinctions are important as different behaviours may influence risk of collision. To estimate turnover for a specific area, we, therefore, need to partition the activity of birds into each category to provide an estimate of turnover specific to each behaviour. Outputs from this project may then be compared to at-sea survey data that classifies observed individuals into these three behavioural categories.

Care must be taken in relating the outputs of this work to at-sea survey data. In this project we estimate turnover in relation to a complete "census snapshot" survey of the footprint, thereby assuming that the number of birds present in the survey area can be known exactly at a particular instant in time. The idea of the "census snapshot" is that we have data on the location of all birds within the survey area ( e.g., the wind farm footprint) at the exact time of the survey, and that we know the behaviour of each of these birds. The survey is, therefore, assumed to be comprehensive ( e.g., a census) and to take place instantaneously (a snapshot). This assumption represents an idealised situation - in reality, survey data will typically not be a complete census (because only part of the population in the area will be counted), and will typically not be instantaneous (it will take some time for the survey to be conducted). The biases associated with the actual observation process (up-scaling, non-detection) are also important in at-sea surveys, and need to be considered when translating survey data into an estimate of the overall population using a site (Thomas et al. 2010). However, none of these factors are directly related to turnover. The key motivation for our formulation as a "snapshot census" is to separate out the quantification of turnover (which is a property of the population itself) from the quantification of observation error (which is a property of the survey method). The latter issue is beyond the remit of this project, and its effect will differ between different survey methodologies. However, it must be considered when relating estimates of turnover to the outputs from at-sea survey data.

When assessing turnover, we consider the following Special Protected Areas ( SPAs): Buchan Ness to Collieston Coast SPA, Fowlsheugh SPA, Forth Islands SPA and St Abb's Head to Fastcastle SPA. Four recently consented wind farms are considered - Neart na Goithe ( NNG), Inch Cape ( IC), Seagreen Alpha (Alpha), and Seagreen Bravo (Bravo; full details are in Searle et al. 2014). To explore the effect of the size of a wind farm footprint on estimates of turnover we also consider 'artificial' footprints of fixed size, centred on the geographical coordinates of the recently consented footprints.

Data on bird distributions for the four species (kittiwake, guillemot, razorbill and puffin) were taken from GPS loggers that had been deployed on individual birds from the four SPAs in the region of the recently consented wind farms during chick-rearing periods in 2010, 2011 and 2012. GPS tracking data enable us to estimate the relative spatial densities of birds that have come from a specific SPA. For each species, bird densities were estimated from the filtered GPS tracking data using a Binomial generalized additive model ( GAM). The GAMs provide an estimate of the predicted bird density (of breeding individuals during the chick rearing period) for each species-by- SPA combination (for more details see Searle et al. 2014). Our approach is based on the use of tracking data from birds of known breeding origin, and would not be directly applicable for sites and species where such data are not available. However, the use of GPS technologies is becoming increasingly affordable, and collecting tracking data to estimate at-sea turnover of individuals from data deficient SPAs should be prioritised in any future research.

We modify the simulation of locations to vary the extent to which birds display site fidelity to foraging locations because we expect this to be crucially important when considering turnover. In so doing we make two simplifying assumptions regarding how site fidelity is expressed by foraging birds. These assumptions are made partly for reasons of computational speed, and partly because of a lack of relevant empirical data to inform more realistic representations of site fidelity. Firstly, we assume that there is complete fidelity within a time-step (24-36 hours depending on the species), such that each bird visits only one foraging location for each day (or 36 hour period) of the breeding season - although the bird may visit this site more than once. The second assumption is that site fidelity operates in relation to the cells of a regular grid (0.5 km x 1.0 km) used to model the density of foraging birds, and not to other spatial areas such as irregular shaped areas that may better represent foraging hotspots.

Finally, we categorise the activity of birds into each behavioural category by using empirical data from bio-loggers for each species to estimate the proportion of time birds spend in each activity during a time period of 05:00 to 20:00 hours. This time period was chosen to coincide with the part of the day over which at-sea surveys typically take place (Camphuysen et al. 2004). We also investigate an alternative methodology in which activity budgets were derived from the outputs of a foraging model developed in a previous project (Searle et al. 2014).