Linking eBird data with high spatiotemporal remote sensing products to estimate occupancy of bird populations across the island of Ireland

The vast increase in biodiversity data generated through citizen science initiatives, alongside a growing suite of remote sensing products and advanced modelling tools, has opened new avenues for rapidly, accurately and efficiently monitoring species trends to inform conservation, management and policy. However, fully harnessing and integrating these diverse data sources and models remains underexplored, with efforts often concentrated in a limited number of developed countries. In this study, we (1) summarized two decades of eBird data across the island of Ireland and (2) combined 4 years of data with high‐spatiotemporal‐resolution remotely sensed information and Bayesian hierarchical occupancy models, to assess species occupancy and identify environmental drivers of occupancy, accounting for imperfect detection. We applied a single‐species multi‐year occupancy model to the Common Blackbird Turdus merula and a multi‐species multi‐year model to 14 species, validating model predictions using a systematic bird survey. Results show an exponential increase in eBird checklist submissions across the island of Ireland between 2002 and 2023. The single species model fitted the data well and accounted for several observation biases, and occupancy predictions resulted in high accuracy with an area under the curve or the receiver operating characteristic (AUC‐ROC) of 95.6% across years. The multi‐species model presented lack‐of‐fit, and model prediction accuracy varied across species and years: 14% were poor (AUC‐ROC < 60%), 34% fair (AUC‐ROC 60–70%) and 77% good (AUC‐ROC > 70%). Although more data are needed to improve fit in the multi‐species model and include rarer species in the analysis, our findings highlight the potential of integrating eBird citizen science data with high‐resolution remote sensing imagery to enhance current conservation, restoration and monitoring programmes. Assuming that citizen involvement and checklist submissions continue their current increasing trends, spatial and temporal data coverage will improve as well as model outputs. To guide this process, we offer guidelines for generating more ‘analysis‐friendly’ checklists based on the results of this study. All code used here is also provided to facilitate replication and adaptation by researchers and practitioners.