IAP-26-086

Global bird community change and its drivers: building reliable evidence from uneven monitoring

How and why have bird communities changed over time? Repeated bird surveys at a sample of locations offer opportunities to answer these questions across continents. But would we get the same answers if we monitored a different set of locations or even the whole planet? For example, surveys concentrated in stable landscapes could miss the consequences of habitat conversion or rapid warming. This can bias perceptions of biodiversity trends and of the effects of global change pressures. Uncertainty estimates from fitted models do not necessarily capture this problem (Boyd et al. 2022).

This project will investigate whether accounting for uneven environmental representation changes our understanding of bird community trends and their relationships with environmental drivers. You will connect bird survey time series with environmental histories across North America, Europe and East Asia, developing approaches that can be applied in different regions worldwide.

The project combines ecological discovery with improving evidence for biodiversity policy. For example, reliable estimates of how and why bird communities are changing are needed to monitor progress towards the Kunming-Montreal Global Biodiversity Framework, while the EU Nature Restoration Regulation similarly sets targets for national common farmland bird indices. Working with researchers at Durham University and the UK Centre for Ecology & Hydrology (UKCEH), you will investigate how evidence used in such assessments could be strengthened.

The project will use repeated bird survey time series, such as those collected by the Breeding Bird Survey across the Northern Hemisphere, to estimate changes in bird communities and drivers thereof. Survey locations will be linked to spatially explicit land-cover, land-use and climate time series. Analyses will adjust the data collected at monitored locations so that it is representative of the wider regions in terms of land use and climate. Initial work will establish an analysis protocol for a manageable set of core datasets before upscaling the approach to additional regions. Four connected components will guide the research.

1) Assess representation:
Quantify how environmental-change trajectories at locations with bird community time series differ from those across the broader regions they are used to describe. For each analysis, define the target region, observation period, and the ecological quantity to be estimated, and compare monitored and regional environmental trajectories at corresponding spatial and temporal scales. Identify which types, magnitudes and combinations of land-use and climate change are overrepresented, underrepresented or absent from the available observations, and assess how this representation changes through time as monitoring locations enter or leave the dataset. Compare these patterns across regions and monitoring datasets to establish where the evidence is environmentally representative and where it is not.

2) Assess consequences:
Investigate how the locations selected for monitoring and their uneven representation of environmental change influence estimates of bird community change and environmental driver effects. Simulate virtual worlds in which bird communities and their environmental drivers change through time and across space, using real monitoring data both to make the simulations as ecologically realistic as possible and to complement their findings. Combine simulations with resampling experiments and empirical comparisons. Known simulated trends and driver relationships will provide benchmarks against which estimates can be evaluated, while empirical analyses will assess sensitivity in real monitoring systems. Identify the circumstances in which conventional ‘off-the-shelf’ analytical approaches produce poor estimates. Determine when uneven representation alters the estimated magnitude or direction of community trends or changes conclusions about their drivers, and which conclusions remain robust. Examine how these effects depend on the environmental changes sampled, ecological responses and monitoring coverage, while distinguishing bias arising from site selection and environmental representation from confounding and other sources of uncertainty.

3) Develop and test methods to reduce representation bias:
Using the simulations described above, develop and test alternative approaches that account for uneven representation of environmental-change trajectories in estimates of bird community trends and environmental driver effects. Compare weighting and model-based approaches where appropriate, drawing on recent work linking sampling theory, missing-data methods and causal inference (Boyd et al. 2025). Explicitly assess the limits on reliable community-change and driver-effect estimates imposed by environmental conditions that are poorly represented or entirely unobserved, for which weighting alone cannot recover missing information and extrapolation requires additional assumptions. Evaluate sensitivity to these assumptions and identify where inference should be restricted.

4) Improve biodiversity-change inference and indicators:
Apply the resulting approaches to regional indicators of bird abundance or occupancy change and to estimates of environmental driver relationships, testing their reliability across contrasting regions and distinguishing uncertainty conditional on fitted models from limitations arising from representation and extrapolation. The project will use these findings to develop guidance for biodiversity status-and-trend assessment.

This project will be supervised by Carsten Meyer at Durham, with Rob Boyd and Gary Powney at UKCEH as co-supervisors. Besides regular online meetings, optional research visits to UKCEH will support collaboration. Through UKCEH, the student will have opportunities to engage with policy makers and those involved in the production of national biodiversity indicators, thereby enhancing the impact of this project by informing indicator development and guidance on interpretation. Specific analytical choices will be refined with the student.

Year 1

Complete initial training and literature review; select core survey datasets and target regions; define inferential goals and harmonise bird and environmental data. Undertake the first representation assessment and begin writing it up. Start discussions with indicator producers about useful outputs.

Year 2

Complete and write up representation analyses and submit your first paper. Develop simulations and resampling experiments to assess consequences for community trends and driver estimates and write it up. Begin developing and testing adjustments alongside these experiments, prioritising the most consequential sources of bias.

Year 3

Complete and write up analyses of consequences of representation bias and submit your second paper. Evaluate adjustments across additional regions and monitoring contexts. Produce and compare adjusted and conventional indicators and driver estimates, with sensitivity analyses. Discuss interpretation with indicator producers, prepare guidance and your third manuscript, and draft thesis chapters.

Year 3.5

Complete cross-region synthesis, finalise and revise manuscripts, and submit your thesis within 3.5 years. Release documented analytical code and share findings and indicator guidance with research and policy audiences.

Training will develop skills in and knowledge of reproducible programming, spatial environmental-data integration, ecological statistics, sampling and missing-data theory, causal inference, and environmental policy. The supervisory team will support analysis of bird community time series and development and communication of biodiversity indicators. IAPETUS training will complement project-specific learning with training in scientific writing, presentation and engagement with evidence users. Training and the balance of ecological and methodological work will be tailored to the student’s background and interests, building expertise relevant to research, conservation and environmental assessment.