Lack of sound science in assessing wind farm impacts on seabirds

Abstract

Electrical power generation from wind farms has grown rapidly in the UK and European Union (EU) in the last decade and is set to grow further. By 2020, the EU proposes to source 20% of energy from renewable sources (Directive 2009/28/EC). Wind energy is expected to provide 9–14% of global electricity generation by 2050 (IPCC 2011). This may eventually reduce climatic change and its negative impacts on biodiversity, but there are also several poorly quantified negative effects on wild species of renewable energy generation, including wind turbines. For example, birds and bats are killed by colliding with turbine blades or towers and there may be effects of wind farms on mortality and reproductive rates of a wide range of species from avoidance and displacement. Birds may incur additional costs or forego benefits because of reduced transit or foraging within or near to wind farms (Drewitt & Langston 2006; Searle et al. 2014). Depending upon the strength of density-dependent compensatory processes, these effects could reduce the population to a lower stable level or cause its extinction (Wade 1998; Niel & Lebreton 2005). Except in the rare circumstances where density dependence is exactly compensating, such effects would always diminish population size. Positive effects of renewable energy infrastructure on populations of wild species have also been proposed and, in a few cases, quantified. These include possible enhancement of food resources of seabirds by protection from fishing from the presence of offshore installations and the provision of artificial substrates as habitat for fish and invertebrates (Inger et al. 2009; Langhamer, Wilhelmsson & Engström 2009). The UK has the best wind resources in Europe (DECC 2011). Although the cost per megawatt-hour of electricity generation from offshore wind turbines averages about twice that for onshore installations (Bilgili, Yasar & Simsek 2011; Chu & Majumdar 2012), offshore wind power is currently favoured over onshore by the present UK government because of public perceptions of nuisance and landscape consequences of onshore turbines. The UK also has internationally important breeding populations of seabirds. It holds more than 10% of the world's breeding population of eight species, of which three have more than half of their global breeding population in the UK (Brown et al. 2015). Because seabirds range over long distances, there may be cumulative impacts on a breeding colony from several wind farms (Masden et al. 2010). Seabirds are long-lived and late-maturing, which renders their population growth rate particularly sensitive to additional mortality from collisions or displacement (Niel & Lebreton 2005). The importance of these seabird populations and their sensitivity places a heavy responsibility on those conducting and acting upon scientific assessments of the impacts of offshore wind farms on seabirds to comply with the protection measures and the precautionary principle enshrined in the EU Birds and Habitats Directives (Directive 2009/147/EC and Council Directive 92/43/EEC). For the UK, and other countries within the European Union, the regulation of wind farm construction requires the assessment of possible damage to the integrity of sites and populations under the EU Habitats and Birds Directives. Consideration must be given to impacts on bird populations of a project on its own and in combination with others already in existence, given consent or planned. Governments give or refuse consent for the construction of wind farms after taking into account the scale and level of certainty of the impacts indicated by these assessments. However, there are no definitive quantitative thresholds or criteria defining how large or likely expected impacts must be for damage to the integrity of sites and populations to be anticipated and for consent for wind farm construction to be denied or limited. Consent can be granted only if it is ascertained that there will not be an adverse effect on the integrity of a Natura site, excepting in cases where there are imperative reasons of overriding public interest for consent and no alternative solutions (Article 6(4) of Directive 92/43/EEC). In recent years, several plans for large offshore wind farms have been approved and some built in UK and EU waters close to large seabird populations because the competent authority judged there was no expected adverse effect on the integrity of the Natura sites involved. For example, in 2014 approval was granted for several extensive wind farms at Hornsea (England, UK Government) and the Firth of Forth (Scotland, Scottish Government), close to internationally important breeding populations of seabirds. This approach contrasts with that in some other EU states. In Germany and Denmark, for example, offshore wind farms have been subject to rigorous marine spatial planning with the aim of avoiding potential conflict with nature conservation as part of the required Strategic Environmental Assessment (SEA) process recommended in EU Commission guidance (European Commission 2011). The German Cabinet approved Europe's first maritime spatial plan in September 2009, after a considerable effort in terms of surveys and research to identify marine sites of high nature value and potential conflict areas with wind farms and to establish zones for various activities and infrastructure. The offshore SEA covering UK waters is not of comparable quality. In this perspective, we argue that the methods and data used in these cases for estimating effects upon seabird demographic rates and translating them into potential impacts on seabird populations do not allow adequate assessment of effects on site integrity. As a result, sound science and its logical interpretation are lacking in Environmental Impact Assessments of this large and expanding industry. Collision risk models (CRMs) are used to predict the number of fatal collisions of flying birds with wind turbines and per capita additional mortality rates. In the UK, the most widely used CRM is that of Band (2012) (see review by Masden & Cook (2016)). The model requires estimates or assumptions about bird numbers and ages at the wind farm, attribution of birds at the wind farm to source populations, sizes and age structure of source populations, flight behaviour and avoidance rates. Data specific to the project and species being assessed are usually collected on seabird numbers and flight heights, judged by eye, but these estimates are subject to substantial uncertainties, variability and potential biases (Johnston et al. 2014), including: Total avoidance rates used for CRM calculations for seabirds, including within-wind farm avoidance of individual turbines and macro-avoidance by movement of birds around the turbine array, are most often based upon judgement or extrapolation from other contexts rather than pertinent data. Empirical values are only available from a few species (mostly gulls and terns) and usually extrapolated from studies of onshore wind farms, where different circumstances prevail (Cook et al. 2014). Robust direct estimates of within-wind farm avoidance rates are lacking for seabird species frequently present in and near planned and consented offshore wind farms in the UK, such as northern gannet Morus bassanus and black-legged kittiwake Rissa tridactyla (Cook et al. 2014). Macro-avoidance and displacement rates have been estimated using radar, visual surveys and imaging, but robust quantitative estimates with confidence intervals are generally not used in impact assessments. Estimates of macro-avoidance for the same species can be highly variable (e.g. Petersen et al. 2006; Krijgsveld et al. 2011; Vanermen et al. 2012, 2013 for northern gannet). This may well be because macro-avoidance varies with the relative positions of nesting and foraging sites, foraging site quality and seasonal timing of studies. At onshore wind farms, carcasses of some of the birds killed by collisions with turbines can be collected during systematic searches and probabilities of their detection can be estimated. This allows estimation of numbers of deaths per unit time and confidence intervals, even if with low precision (e.g. Bellebaum et al. 2013). These methods help to quantify uncertainty and remove bias, but are currently impractical for offshore wind farms. Alternatives that use video or thermal camera systems have not yet been deployed sufficiently to substitute for them. Where direct measurements of avoidance rates are lacking, Band (2012) recommends use of a range of plausible values. However, this can result in a 20-fold variation in assumed per capita mortality rates (APEM 2015). Overall, CRM outputs are sensitive to the combined effects of multiple assumptions of unknown accuracy, sampling errors and unquantified biases. Only for species that almost completely avoid entering wind farms can the annual per capita mortality rate from collisions be estimated reliably and with robust confidence limits (Desholm & Kahlert 2005). Validation tests of offshore seabird CRM outputs, in which expectations from pre-consent data and modelling are compared with independent robust post-construction measurements of numbers of collision deaths, have not been conducted. Estimation of effects on seabird demographic rates of the displacement and barrier effects of wind farms is even less well developed. Avoidance of wind farms by foraging and migrating birds can be substantial and operate over long distances from the turbines (Desholm & Kahlert 2005; Petersen et al. 2006; Percival 2010), but the degree to which this affects travel times and costs, access to food and mortality and reproductive rates of breeding seabirds has not been measured reliably. In the case of migrating birds, the displacement and increased travel costs caused by avoidance of a single wind farm may be trivial relative to the total length and cost of the journey (Masden et al. 2009), but effects on demographic rates have not been robustly quantified by empirical studies for central-place foraging breeding seabirds repeatedly subjected to barrier or displacement effects. Simulation modelling has been performed of potential effects of displacement by as yet unconstructed wind farms on seabird time and energy budgets and demographic rates (Searle et al. 2014). Modelled potential effects of displacement included considerable declines in adult survival of up to 2·1% for black-legged kittiwake and up to 4·9% for Atlantic puffin Fratercula arctica (both for the Forth Islands cumulative effects: table 3·3 of Searle et al. 2014), though simulated effects on survival for other species and sites and for breeding productivity generally were small. The species for which collision mortality can be reliably estimated as low, because of strong avoidance, are those for which displacement and barrier effects upon demographic rates are potentially the largest, but currently unquantified. In summary, the procedures currently used to calculate expected effects of proposed wind farms on seabird per capita mortality rates and breeding success largely involve modelling with little firm empirical data. Moreover, actual outcomes at wind farms that have been constructed have not been measured, so model predictions are not tested and there is no adaptive improvement of the decision-making process (Nichols et al. 2015). As a result, scientifically robust and defensible calculations of effect sizes for changes in seabird demographic rates caused by collision, displacement and barrier effects of offshore wind farms, with confidence intervals, are currently lacking. Assessments of the impacts of offshore wind farms in the UK on seabirds require that the highly uncertain estimates of effects on demographic rates are translated into projections of impacts on population size or trend. Decisions about UK offshore wind farms have been based upon, or influenced by, the following effect–impact translation procedures. The recommended and robust application of this method is to identify a level of additional mortality above which a decline of the affected population to eventual extinction would be likely (Niel & Lebreton 2005). In recent cases, such as Hornsea, the UK statutory conservation agencies advised using this method in wind farm assessments to identify demographic rate thresholds below which additional mortality estimated from CRMs and related methods is unlikely to adversely impact the population (Natural England 2014). This reverse application involves faulty logic because PBR's value of maximum potential excess growth may not be realizable in the ecological circumstances of a particular population of interest. In addition, PBR does not estimate the effect of additional mortality on population size. Potential biological removal provides thresholds of additional mortality that are sensitive to assumptions made about the form of density dependence. The studies of Wade (1998) and Bellebaum et al. (2013) show that the shape parameter of the generalized logistic equation has a strong effect on PBR results. Details of the form of density-dependent relationships are rarely known for animal populations and are unknown for any of the UK seabird populations to which PBR has been applied. These uncertainties have prompted the use of ‘recovery factors’, which are constants by which the maximum possible value of the PBR threshold is multiplied to give a safety margin (Dillingham & Fletcher 2008). The values used for these recovery factors are based upon judgement. There has been no empirical validation of their safety by observation of the effects on population size of known additional mortality rates from any source in any bird species. This method, which has not yet been published in the peer-reviewed scientific literature, was developed by Marine Scotland, a Scottish government agency, and used in a recent assessment of the impact of wind farms on internationally important seabird populations in the Firth of Forth (Marine Scotland, 2015). It uses probabilistic forecasts from stochastic seabird population models to assess the probability of a particular level of population size occurring at some future time, such as the end of the period of operation of a wind farm, in the absence of the wind farm. In practice, this probability is obtained from a simulation model of the population in which variation in expected future population size arises from supposed future demographic and environmental stochasticity in demographic rates, when to the population of a size over a period of years, which is the period for an offshore wind farm. the best estimate of future population after the expected effects of the wind farm on demographic rates are into or the population size that is likely to be or in the absence of the wind farm, that the impact of the wind farm is The of this approach are the of projections of the demographic rates used in the model of the seabird population long into the future is highly uncertain and the the estimated the the population it does not the uncertainties in size of the effects of the wind farm on demographic rates, which are unquantified. does not assess the risk or probability that the wind farm will cause a particular or change at It proposes that an half as likely to as not if there is no wind farm be the threshold for the threshold probability for is and is from an about the to the of an or of at given size based upon available et al. 2010). The threshold for is as in the However, this was not developed for the of of which also requires that the costs and benefits of possible outcomes are It is not only the of being that is but also the scale of the damage caused by being is given by the of for using as a risk threshold for damage to important nature conservation sites and their species a from a developed by for a different uncertainties in future seabird population changes not if in the probability of a specific population with and wind farms could be reliably and used as criteria for This on in risk has been proposed by the & England It was that assessments of impact be based upon an threshold level of the and in the probability that a population decline by an of the level would In this approach is to because it the uncertainty in the of the effect of the wind farm into However, the of this are sensitive to the of demographic rates. For example, in a model in which the values of demographic rates a in population it is unlikely that even large additional mortality would give to an in the probability of population decline and probabilities would be small. the rates were and the values to the population being the same level of additional mortality could result in a large in the probability of population decline and In practice, uncertainties in future projections of and populations are so the probability of an for population size be This such as and which are based upon assessments of probability or in probability given present the effect–impact translation procedures above have a threshold for an thresholds are to because to a to establish damage to the integrity of a site will or will not However, in the case of and the thresholds only because are have no in population and the of some adverse impact on population size. PBR does identify a threshold based upon population it is that is to the at PBR could be used to identify a threshold level of effect of wind farms on demographic rates above which a decline of an affected population to eventual extinction would be almost However, population declines of a wide range of of could be caused by effects of wind farms on demographic rates well below large these declines would be upon the form and strength of density which are unlikely to be measured with and the of such declines has not been quantified using PBR in any UK wind farm argue that such declines would adverse effects on site integrity. PBR is not an method for population impacts of a in a that is to the of the public and more robust for impacts of wind farms on seabirds is to using a model expected population with and the expected effects on demographic rates of the wind farm, at the end of its The of the expected population size with the wind farm to that it of population is a robust for likely impact of a set of effects of the wind farm on seabird demographic rates. This approach is to the assumptions made about the variability and of demographic rates in the model from which it is because the same uncertainties to the and this effect–impact translation little to the uncertainty in the in population size caused by the wind farm. dependence to reduce the impact on population size of a given effect of the wind farm on demographic rates, so the from the model is a precautionary it that density-dependent in UK seabird populations and that including it in (e.g. & could to more estimates of population impact than those based upon However, would only be increased if robust estimates of the form and strength of density dependence were available or population outcomes could be to be to assumptions made about density dependence in the absence of In practice, no assessments of population impacts of additional mortality from wind farms on UK seabirds have included empirical estimates of the form and strength of density dependence because estimates not to be adequate of density dependence is we the use of density dependence is included or there is no threshold value of built into the estimates from for and population models using a approach (Marine Scotland, have been as part of offshore wind farm impact but their have not been used as in decision-making about the or of UK offshore wind farm upon the of UK wind farm we that methods such as and have been used in to because provide thresholds which can be used to argue that site integrity will not be affected by the do not provide a argue because the thresholds by the other methods are and be used as the method, and we that any of the potential effects of offshore wind farms on seabird demographic rates, if could be included in an assessment using an procedures for empirical modelling effects on demographic rates and translating those effects into impacts of offshore wind farms on seabird populations are Empirical measurements of effects of offshore wind farms on seabird demographic rates from are not sufficiently and In the case of some important such as turbine avoidance rates and the strength of density-dependent estimation is rarely even As a result of these in the the of effects of wind farms on seabird demographic rates be estimated and the level of and precision in the estimates used be these require the renewable energy to an adequate level of research to estimate effects of wind farms on seabird demographic rates more reliably. The to this but the of its project to be with to the number of species to their breeding and of and of radar, and are likely to be including et al. 2015). defensible approach is to these effect and their uncertainties, into expected impacts on that the population from a model would be an method for this from these of estimation and there is a logical in the scientific assessment and decision-making about the of wind farm have been that to an threshold for a negative impact of a wind farm on seabird populations, below which this negative impact is as no adverse effect on site integrity. However, the has no the thresholds used to the of offshore wind farm impacts are poorly not for the and have no biological it is to decision-making of translation is At data are being combined with and scientifically thresholds to argue that wind farms will cause no damage to the integrity of sites to and Europe's that the probability of of an animal population and its time to extinction generally with its size & to European Commission guidance on the of sites by the EU Birds and Habitats Directives of the Habitats Directive provides that integrity of the site involves its ecological The as to it is adversely affected on and be to the conservation (European Commission In addition, for the integrity of a site not to be adversely a of of the European Union of of the European Union that the to be at a conservation which of the of the site that are to the presence of a habitat was the the of that upon this we argue that some damage to the integrity of a site will have been if populations of the seabirds for which it was are even to a by the effects of a wind farm, compared with would have that is expected to be the it does not that the competent authority give consent for a wind farm. 6(4) of the Habitats Directive tests that the expected damage can be and However, science not be used to avoid those tests by that no damage will Data have not been because this does not data. and are conservation at the for the of Birds is the environmental are in the assessment of the impacts upon birds of proposed wind farms in the