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

2 Introduction

The Scottish Government has set a target to generate 50% of Scotland's overall energy consumption from renewable sources by 2030 and to have decarbonised the energy system almost completely by 2050. The marine environment offers considerable potential for renewable energy, through wind, wave and tidal stream energy generation. The Scottish Government has a duty to ensure that offshore renewable developments (ORDs) are achieved in a sustainable manner, by protecting the natural environment from adverse impacts in accordance with the Marine Strategy Framework Directive (2008/56/EC), the Habitats Directive (92/43/EEC) and the Birds Directive (2009/147/EC). To help achieve this, they are required, under the Precautionary Principle, to ensure that decisions are informed by the best available scientific evidence and make reasonable effort to address any gaps in knowledge. Crucially, offshore renewable developments have the potential to impact on seabird populations that are protected by the EC Birds Directive, notably from collisions with turbine blades and through displacement from important habitat (Drewitt & Langston 2006; Larsen & Guillemette 2007; Masden et al. 2010; Grecian et al. 2010, Langton et al. 2011).

The process whereby seabirds collide with wind turbines is assessed using collision risk models (CRM). These are relatively simple mechanical models that estimate the likelihood of a bird of a specified size and speed colliding with a rotating turbine of a specified size and speed (Band 2012). These data are combined with densities of birds in flight and their flight height, which are estimated from baseline characterisation surveys of proposed wind farm projects and from generic datasets (particularly for flight heights). The first offshore CRM (Band 2012) was deterministic and provided monthly predictions of species-specific collision mortality for four model options. Two model options (Options 1 and 2) use the 'basic model' and make the assumption that seabirds were evenly distributed across all heights of the wind turbines, Option 1 using a site-specific estimate of the proportion of birds at collision height and Option 2 a generic estimate derived from analysis of a pooled dataset (Johnston et al. 2014). This assumption is known to be incorrect, because the relative frequency of flight densities declines with increasing height. This is important because the risk of collision is not the same for all distances from the rotor hub and therefore varies with flight height, even within the rotor-swept zone. The third model option (Option 3) takes account of this predictable decline in flying densities with increasing height, but uses the generic flight height distributions in Johnston et al. 2014. Option 4 allows the use of site specific flight height distributions, but is otherwise identical to Option 3. In order to improve on the deterministic outputs from this model subsequent stochastic versions of the Band model (2012) have been developed (Masden 2015, McGregor et al. 2018). These models translate the uncertainty and variability in the input values through to outputs that provide predictions, together with associated probability distributions or confidence intervals. Part of the Band (2012) and McGregor (2018) model incorporates a correction for differences between observed and predicted collision rates. This correction factor is usually called the 'avoidance rate' because it is assumed that the majority of this difference is due to the avoidance behaviour of birds in relation to the presence of the wind farm, wind turbines and turbine blades (because model input data are collected prior to windfarm construction and, therefore, do not include any potential displacement or avoidance effects). However, as this correction factor is derived from a collision rate predicted from the Band (2012) model in the absence of avoidance behaviour, it also incorporates model error (currently estimated at ± 20%) arising as a result of simplifications within the model (Band 2012). A key source of this error is likely to relate to how the flux rate is calculated, which means that a greater number of birds may be exposed to the risk of collision than is realistic (Masden et al. in prep). As a consequence, it is important that any avoidance rate used by the Band (2021) model corrects for this simplification in how the flux rate is estimated.

At present, the avoidance rate is a species, but not site, specific value. The overall avoidance rate is often calculated from three different spatial scales of behaviour: macro-avoidance (birds avoid entering the wind farm at all), meso-avoidance (birds enter the wind farm, but avoid entire turbines) and micro-avoidance (birds approach turbines but avoid colliding with the moving blades). Macro-avoidance behaviour has been assumed to be comparable to displacement behaviour, but there are subtleties in the assumptions and interpretation of how these processes are represented by parameters within alternative modelling methods, and in how they are applied to data. Not least, macro-avoidance behaviour only applies to birds in flight, while displacement also applies to birds on the water.

Displacement impacts during the breeding and chick-rearing seasons are currently assessed using two general methods; individual-based simulation models, and simple matrix approaches. Individual-based models predict the time/energy budgets of individual animals and translate these into projections of demographic rates, such as adult annual survival and productivity (e.g. Searle et al. 2014, Warwick-Evans et al. 2017, Searle et al. 2018). These models simulate foraging decisions of individuals under the assumption that they are acting in accordance with optimal foraging theory. Foraging behaviour of individuals is driven by prey availability, travel costs, provisioning requirements for offspring, and behaviour of conspecifics. Impacts of displacement and barrier effects arising from specific OWFs can be assessed using such models by comparing baseline scenarios with scenarios containing one or more OWFs. The models can estimate the change in productivity and adult survival between the baseline and OWF scenarios, the latter process resulting from estimates of adult mass at the end of the breeding season. The most comprehensive model available for estimating the population level consequences of displacement and barrier effects for seabirds is the SeabORD model, developed as part of Marine Scotland's project "Finding out the Fate of Displaced Birds (CR/2015/19)" (Searle et al., 2014, Searle et al. 2018; Mobbs et al. 2018). Although this model was developed within the context of assessing displacement impacts of OWFs during the chick-rearing period, it is a very general model of seabird foraging and provisioning behaviour, and time-energy budgeting. This flexibility, and its ability to track individual birds, makes it ideal for integrating mortalities arising from the different risk types – collision and displacement.

The second method for quantifying impacts of displacement is the Displacement Matrix (JNCC, 2022). This takes a more simplistic approach by multiplying the observed population within a proposed wind farm site by the percentage expected to be displaced and the percentage of those expected to suffer mortality as a consequence of displacement. These percentages are currently most often derived from expert opinion (JNCC 2015, JNCC 2022). Displacement in this context is an integrated estimate of displacement (loss of access to habitat) and barrier effects (additional flight costs incurred flying around OWFs), but the method does not separate out rates or mortality consequences into these two categories. In acknowledgement of the uncertainty in both these percentages, a wide range of displacement mortality values is typically presented in impact assessments, potentially using a matrix form with 0-100% displacement along one axis and 0-100% mortality along the other, although in practice the range for each axis is much less than 0-100% (JNCC 2015; JNCC 2022)

The matrix approach and SeabORD have different data requirements, and currently are applicable to different suites of seabird species. SeabORD is more data intensive, requiring utilisation distributions (UD) for each breeding colony and a prey availability map, and is currently only parameterised for four seabird species (common guillemot, razorbill, Atlantic puffin and black-legged kittiwake). The matrix method requires estimates of densities of birds within footprints only (not a full UD), and is applicable to any species for which these data are available. Both modelling approaches require users to specify rates of displacement from footprints, usually derived from expert elicitation, and rarely from empirical studies as very few suitable studies have been completed. The matrix approach also requires users to input a mortality rate for displaced birds, derived from expert opinion, whereas this is a model output for SeabORD.

Currently, collision mortality is estimated separately from displacement and barrier effects, and the effect sizes are combined in assessments of effects of OWFs on seabird populations for those species where both effects are assessed. However, there is concern with this approach because the parameters used in both approaches are not equivalent, which means the resulting estimates are not based on an internally consistent set of assumptions. For instance, the collision model parameter for macro-avoidance appears to be aimed at capturing the same process as the displacement rate parameter in displacement models; however, due to the inclusion of the avoidance rate correction in collision models, it is not clear that these two parameters are in fact directly comparable. Second, there is concern about double counting: an individual seabird is potentially vulnerable to displacement and collision, yet it can't be vulnerable to both simultaneously. There is, therefore, a concern that assessments may be overestimating effects on demographic rates due to double counting — that is, counting the death of an individual bird twice due to a failure to separate collision and displacement effects.

Accordingly, the objective of this project was to develop a framework within which collision, displacement and barrier effects can be aggregated into a single overall assessment of combined impacts in a way that is internally consistent, scientifically defensible, and practically useful (sensu Humphreys et al. 2015).