Offshore wind developments assessment - seabird collision risk, displacement and barrier effects: study

7 Wider recommendations

We have identified six main areas for wider recommendations for future research to facilitate robust and defensible estimation of combined collision and displacement/barrier effects for seabirds:

• Requirements for validation of individual-based models;
• Improvements to incorporation of uncertainty in individual-based models;
• The potential for making collision risk modelling an agent-based approach;
• Better unification of the behavioural definitions used within current individual-based models (SeabORD) and those derived from at-sea survey data;
• Potential for combined modelling of displacement and collision risks throughout the whole year;
• Potential for assessing in-combination and cumulative effects of multiple OWFs.

7.1 Requirements for validation of individual-based models

Fundamentally, validation of the mechanisms within individual-based models such as SeabORD requires empirical data on the behavioural response of individual birds to OWFs, and the consequences of those behavioural responses in terms of changes to body condition and demography. These data are best acquired through GPS tracking of individual breeding birds before and after construction, with simultaneous measurements of individual condition, breeding success and survival.

7.2 Improvements to incorporation of uncertainty in individual-based models

Most of the model parameters in SeabORD, even when derived from empirical data, are currently utilised within the model as a single value, such as the mean of a sample of empirical data, or as a single value derived from expert judgement. Primarily, this is due to a lack of empirical data with which to test some of the key mechanisms within the model to do with bird behaviour and responses to OWFs, such as intake rates of individuals in relation to prey availability, competition effects of conspecifics on individual intake rates, probabilities of chick mortality arising from adult unattendance, and other such mechanisms. Before such empirical data become available for parameterising some of these mechanisms, the model could use simulated values across a range of plausible values derived from expert elicitation for these unknown parameters to better incorporate uncertainty in model processes within the simulations. This would require improving model processing time efficiency, to allow for a large number of simulations to be run, and to allow for an initial model sensitivity analysis to be conducted to identify key model parameters having the greatest impact on model output. Then, as empirical data is collected, it can be used to refine parameter estimates in the model, thereby reducing uncertainty in model outputs.

7.3 The potential for making collision risk modelling an agent-based approach

Developing individual-based models for collision risk modelling would improve estimates for collision mortality by allowing more nuanced simulations of the response of individual birds to OWFs and turbines, and facilitating better quantification of avoidance behaviours at the three main scales (macro, meso and micro). This could be achieved by developing 3D movement models derived from GPS tagging data, allowing for simulations of individual bird flights to be performed within models. This would also allow a full integration of displacement, barrier and collision risk modelling within a single modelling framework, reducing the need for harmonising parameters across alternative modelling approaches, minimising error, and facilitating quantification of uncertainty. This development would also allow for a better representation of bird movements in relation environmental parameters (e.g. wind speed and direction) and diurnal patterns. This will require the collection of high frequency GPS tagging data of individuals interacting with operating OWFs, preferably across many individuals and multiple colonies to appropriately capture individual variation in behavioural responses.

7.4 Better unification of the behavioural definitions used within current individual-based models (SeabORD) and those derived from at-sea survey data

The SeabORD model simulates a range of individual behaviours, with individuals at-sea potentially performing commuting flight, foraging or resting at sea. These behaviours are identifiable from GPS tracking data, particularly when coupled with other technology such as TDRs or accelerometers (e.g. Thaxter et al. 2019). However, at-sea survey data tends to only be classified into individuals in flight or on the sea surface (though some other behaviours may be observed, such as multi-species aggregations of birds), meaning that estimating utilisation distributions (UDs) from at-sea survey data cannot currently provide foraging-specific UDs, and also requires the use of apportioning methods to generate UDs for specific breeding colonies. Other behaviours that are of interest in understanding the use of a proposed wind farm can be obtained from at sea survey data, but not from GPS tagging data (such as density surface models of all birds using the space, or the spatial occurrence of multi-species feeding assemblages). The estimation of behaviour-specific UDs to facilitate modelling methods such as individual-based models, or indeed to refine estimates used in sCRM or the matrix approach, is therefore best approached through the use of GPS tagging. However, at-sea survey data can be used to provide context for GPS tagging data. It is important to recognise that to date GPS tagging tends to sample breeding adults, but assessments require knowledge about all of the birds present in the areas of OWFs (all age classes and non-breeders). At present, at-sea surveys are best placed to provide this categorisation of individuals.

7.5 Potential for combined modelling of displacement and collision risks throughout the whole year

Sub-lethal effects of offshore renewable energy developments on seabirds during the non-breeding season are currently poorly understood, poorly estimated, high in uncertainty and low in defensibility. We, therefore, recommend research to develop a new individual-based model of birds in the non-breeding season, parameterised with at-sea survey data and tracking data on winter distribution and activity budgets of birds from multiple colonies. A similar approach to that of the breeding season model SeabORD could be developed, whereby a simulation model predicts the time/energy budgets of seabirds during the non-breeding period and translates these into projections of adult annual survival and productivity.

There are critical differences between the constraints and behaviours within which seabirds operate in the non-breeding season, compared to the breeding season. In the non-breeding season, adult birds are independent of offspring and mates, and are not typically operating out of a central place such as a breeding colony. Some species undertake partial or full migration in winter, and winter places higher energetic costs upon birds. However, birds are constrained during the non-breeding season by a number of mechanisms, for instance they may be spatially constrained because of high flight costs or physiological changes such as moulting and may be temporally constrained by shorter day lengths. Some progress has been made in developing year-round models for assessing displacement effects of OWFs; van Kooten et al. (2019) developed modelling methods to consider effects for the full life cycle of several seabird species in the North Sea, focusing on the wider population in the region, rather than on specific breeding colonies. Whilst representing an important step forwards, this approach does not yet include reproduction or density dependent effects arising from reduced carrying capacity as individuals are displaced into smaller areas; and was also restricted to conducting separate simulations for the breeding and non-breeding seasons in species present in the region year-round (van Kooten et al. 2019).

Some of the key research priorities for developing such a model include advancing estimates for the extent of interaction between birds of known provenance (SPAs) and OWF footprints during the non-breeding season (GPS tagging); developing methods for defensible estimates of mortality rates of displaced birds during winter (energetic models and changes in mass and survival), and the incorporation of carry-over effects from winter displacement or barrier effects on subsequent breeding efforts.

There is currently the strongest potential for developing such an individual-based model for common guillemots, although this species is thought to be less affected by collision due to the majority of its flight occurring below rotor height. In this species, there exists year-round data from GLS (light-level global location sensor/gelocator) tagging on their wintering distribution, activity budgets and energetics, and subsequent changes in body mass. Other species should also be prioritised by developing GLS deployments in new species to build baseline individual-based models, as well as GLS deployments in new species to estimate interactions with OWF developments.

It is, however, important to understand the legal frameworks in which impact assessment for OWF occurs. These individual based models are likely to be of great utility to estimating year-round impacts on birds from important breeding colonies designated as SPAs. While this fulfils the requirement to assess impacts under the Habitats Regulations (referring to marine SPA sites), it is also necessary for OWF developments to assess their impacts under the relevant EIA legislation. To undertake an EIA the applicant must address impact to all birds potentially affected by the development, not just those that can be tracked as breeding adults from SPA colonies.

7.6 Potential for assessing in-combination and cumulative effects of multiple OWFs

Through the use of individual-based models, in-combination cumulative impacts of multiple OWFs are easily quantified, because the interaction of each individual bird with each OWF is captured within the modelling process. By embedding collision risk models within individual-based models, in-combination assessments are easily made, with no risk of double counting of collision mortalities arising from turnover of observed birds within different footprints.