Abstract
Offshore wind farms are a vital part of the transition to clean energy, but they can affect seabirds by causing them to avoid or be attracted to the wind farm area — a process known as distributional change or displacement. When regulators cannot be certain how much displacement is occurring, they must assume the worst, which can stall or complicate the consenting of new developments. The ImpUDis project was established to address this problem directly. Working with industry, regulators, and conservation bodies, the project reviewed the available evidence, developed a standardised set of methods for measuring displacement, tested those methods on real monitoring data from eight UK offshore wind farms, and engaged key stakeholders to validate the approach. Its central finding is that while displacement is real and measurable, the quality and consistency of monitoring data has historically been the main barrier to confident conclusions — and that addressing this data challenge is just as important as any statistical advance.
A review of 76 studies across 35 European offshore wind farms (WP1) confirmed that seabird distributional change is real but highly variable. Around 41% of studies reported statistically significant displacement or avoidance, with auks, gannets and divers tending to avoid wind farms and kittiwakes showing mixed responses. Reported displacement magnitudes varied enormously between sites and species, reflecting genuine ecological variability as well as major inconsistencies in survey design, spatial scale and analysis method. Critically, a large proportion of existing monitoring datasets were found to be unsuitable for robust analysis, due to missing effort data, inconsistent formats, poor metadata and ambiguous species identification.
Building on this evidence base, a formal guidance framework was developed (WP2) for estimating seabird redistribution in a consistent, transparent and reproducible way. The framework adopts a Before–After–Gradient (BAG) study design — the practical standard for OWF monitoring — and specifies detailed requirements for data structure, modelling approach, and reporting of outputs with quantified uncertainty.
This guidance was tested against real monitoring data from eight UK wind farms (WP3), with approximately 54 redistribution models fitted across sites and species. Results confirmed the framework is broadly workable. Where data quality allowed, functional distance and spatial models were more informative than simple inside–outside comparisons, and the framework produced outputs directly compatible with existing assessment tools. However, a very large proportion of available data could not be used due to the data quality issues identified in WP1, meaning many results must be treated as indicative rather than definitive.
A stakeholder workshop (WP4) brought together developers, regulators, statutory nature conservation bodies, NGOs and academics to scrutinise the methods and results. Participants validated the analytical approach and reinforced the need for better data standardisation, consistent metadata, and more prescriptive survey standards going forward.
The WP2 methodology is designed to be directly compatible with the existing UK consenting framework and can be integrated into assessments in three ways: redistribution estimates with confidence intervals can be applied directly within existing displacement matrices; gradient-based model outputs can underpin a modified matrix approach that distributes mortality spatially rather than applying a single rate; and density surface model outputs can be used as input layers in individual-based models such as SeabORD or DisNBS, enabling more evidence-based counterfactual scenarios.
Adopting this approach in future assessments would be expected to reduce uncertainty significantly. Where current assessments rely on a single precautionary displacement rate derived from the literature, the WP2 framework would instead produce a site-informed estimate with quantified uncertainty and a spatial gradient of effect. In cases where empirical displacement is lower than precautionary assumptions suggest, this is likely to result in more proportionate conclusions in Habitats Regulations Assessments and a reduced likelihood of concluding Adverse Effects on Integrity.
For this methodology to be routinely applied, the following next steps are required: development of a centralised, accessible repository of standardised pre- and post-construction survey data; formal adoption of the WP2 guidance by statutory bodies and industry as the expected standard for redistribution analyses; retrospective data recovery from existing monitoring programmes to expand the pool of usable analogous datasets; and further development of the approach for data-limited species, particularly puffin and divers.
Key recommendations
- The following recommendations are directed at developers, data holders, regulators and statutory nature conservation bodies:
- Future post-consent monitoring surveys must retain and share full effort data — survey tracks and sampled areas — alongside observation records. Without effort data, robust redistribution modelling is not possible, and data stored only as summaries in reports cannot be used.
- All monitoring datasets should adhere to consistent species coding conventions, machine readable formats, defined spatial projections and a comprehensive data dictionary. OWF footprints should be defined using actual turbine locations, not lease areas.
- A coordinated programme to collate existing pre- and post-construction survey data into a shared, accessible repository should be initiated as a priority. This would substantially reduce the data recovery burden in future analyses and enable cumulative learning across sites.
- The BAG modelling framework set out in WP2, implemented through reproducible GAMM/GLMM-based density surface models, should be adopted as the standard approach for future displacement assessments, with outputs including quantified redistribution magnitude and uncertainty suitable for direct input into displacement matrices and individual based models.
- Survey areas should extend well beyond the immediate OWF footprint — up to 10–20 km depending on species — to enable gradient-based models to capture the full spatial pattern of response.
- Regulators and statutory bodies should consider embedding more prescriptive survey and metadata standards in consent conditions, to ensure that future monitoring data are consistently fit for redistribution analysis from the outset.
For more information, read the Literature and data review, Redistribution estimation guidance, Analysis of existing data, and project value statement.