Beta-Diversity Beyond Bias: A Scalable Framework for Reliable Diversity Analysis from Citizen Science Data

Rahil Jasminkumar Amin , Jessie C Buettel, Leon A. Barmuta, Barry W. Brook 

Citizen science data offer unprecedented spatial and temporal coverage for biodiversity research, yet sampling biases compromise their reliability for β-diversity analyses. We introduce a comprehensive framework to address these challenges, integrating space–time scaling, quality thresholds, and multiple partitioning approaches to enhance detection of ecological signals. Applying our framework to 38 million bird observations across southeastern Australia at six spatial resolutions, we demonstrate that uneven sampling distorts β-diversity patterns. Quality filtering altered β-diversity systematically across scales, with reductions ranging from 13% at coarse resolutions to 40% at fine grains. Components related to nestedness decreased more (22–45%) than turnover components (4–30%), indicating that sampling biases primarily inflate richness-difference patterns rather than species replacement processes. Sørensen indices were more sensitive to filtering than Jaccard indices at all scales, confirming theoretical predictions about their differential response to sampling completeness. Local contributions to β-diversity (LCBD) analyses revealed that incomplete sampling artificially inflated community uniqueness measures. This bias could misdirect conservation efforts toward areas that only appear to be biodiversity hotspots due to poor sampling. After filtering, LCBD patterns aligned with known biogeographic boundaries, demonstrating our framework's capacity to recover genuine ecological signals. Our findings reveal a fundamental trade-off: finer spatial grains provide higher resolution but sacrifice coverage, whereas coarser grains maintain coverage but may mask local variation. This scale-dependent framework enables researchers to leverage citizen science data more effectively for β-diversity analysis, ensuring conservation decisions reflect true ecological patterns rather than sampling artefacts.