SMP adopts new models for seabird trends

The British Trust for Ornithology (BTO) released a report recommending new statistical approaches for the Seabird Monitoring Programme (SMP).

It covers 25 seabird species that breed in Britain and Ireland and traces the programme back to its 1986 launch.

The SMP is a scheme that mobilises individuals to monitor breeding seabirds throughout Britain and Ireland on an annual basis, to provide data for the conservation of their populations.

The scheme is funded jointly by the BTO and the Joint Nature Conservation Committee (JNCC), in association with the Royal Society for the Protection of Birds (RSPB), with fieldwork conducted by both non-professional and professional surveyors squad.

Annual data are complemented by periodic censuses, most recently the 2015-21 Seabirds Count.

Trends in seabird breeding abundance and productivity are reported at country levels as Official Statistics in an annual SMP report (e.g. JNCC 2021, Harris et al. 2024).

The production of accurate trends in population studies is important to best inform conservation delivery and policy decision making.

However, when gathering long-term data across a large spatial and taxonomical scale, these data can be inconsistent and/or sparse, creating challenges for calculating population trends.

A recent report to the SMP Steering Committee reviewed the analytical methods used to produce trends for both seabird abundance and productivity for annual SMP reporting.

The review highlighted concerns over whether the current approach used to produce abundance trends (Thomas, 1993) was fit for purpose, particularly due to the high proportion of imputed data used.

The report proposed new methods for obtaining abundance trends, namely four formulations of a Hierarchical Generalised Additive Model (HGAM; Pedersen et al. 2019) and one state-space model (Freeman et al. 2021).

Simulated abundance data were created that matched the spatial and temporal structure of the SMP data for a selection of five species in terms of number of sites, population size strata, missing counts as well as position of missing counts.

These simulations were used to determine how well the proposed methods produced known simulated population trends in the presence of challenges associated with SMP data.

Assessments were made both using only whole colony counts as well as a combination of whole colony and plot counts, sample subdivisions of a larger colony.

Simulation results highlighted the improvements in accuracy when using HGAMs, especially when accounting for stratum level differences and varying site-level trends.

The incorporation of plot count data was also beneficial to provide additional information on the shape of site-level trends.

The state-space model produced similar trends to the HGAM when using simulated data.

The trends produced by the proposed methods were compared to those produced by the current Thomas imputation method when using raw SMP data.

Divergences between the current method and the proposed methods were smaller with SMP data than with simulated data.

In contrast, state-space models applied to SMP data had greater divergence from the proposed HGAM method than was seen with the simulated data, highlighting the need for further refinement of this method by placing additional constraints on the trend and incorporating metadata.

Further consideration of the HGAM approach included an additional simulation study, performed to determine how robust the abundance trends were to mis-specification of strata, as well as simulations to determine if trends can be produced for other species for which abundance trends were produced in the most recent SMP Annual Report, at the UK as well as country levels, with population estimates obtained from the HGAM compared to associated seabird census counts.

An HGAM approach is recommended for production of the SMP abundance trends at this stage, given that observation-level metadata which could inform the detection component of a state-space model are not currently systematically recorded or included in the SMP database.

In the long term, it is recommended that further development of the state-space model would be useful as this would allow for more explicit modelling of observation processes, which is particularly pertinent for species where detection probabilities may be low, e.g. burrow nesters, or situations where observation methods or correction factors are liable to change over time, which cannot be accounted for using HGAMs.

Productivity trends are currently derived from GLMMs fitted using a maximum likelihood method in the proprietary software Genstat.

We updated productivity trend production using a Bayesian framework for GLMMs, which is the recommended framework going forward.

This provided comparable productivity measures to those from Genstat, with the additional benefit of including 95% credible intervals to capture uncertainty in these estimates.

Most uncertainty in these trends was attributed to a lack of data from representative sites across the UK, emphasising the need for more regular monitoring of productivity.

The Seabird Monitoring Programme is funded jointly by the BTO and JNCC, in association with the RSPB, and is supported by Natural England, Natural Resources Wales, NatureScot and the Department of Agriculture, Environment and Rural Affairs, Northern Ireland, and a wider advisory group.

Close collaboration with organisations in the Republic of Ireland enables all-Ireland interpretation of seabird trends.

The findings also highlight the importance of sustained monitoring, better metadata collection, and future development of state-space models to support evidence-based conservation decisions and long-term biodiversity management.

The report is available on the BTO website and can be accessed through the stats page for detailed trend data.

The programme’s monitoring scheme fixtures provides a framework for volunteers to record breeding activity across the country.

The BTO’s commitment to improving statistical methods underscores its role in guiding seabird conservation policy and protected site management.