D5.4 Report of the Indicators of Robustness Assessments

This deliverable documents the conceptual and technical foundations of the dubicube R package, a core component of the Biodiversity Building Blocks for Policy (B3) project. As biodiversity data are often fragmented and heterogeneous, transforming raw observations into data cubes to calculate reliable indicators requires a rigorous assessment of robustness. Here, robustness refers to the stability and reliability of results under variations in underlying data and analytical choices.

The report details an integrated evaluation framework organized into three complementary pillars, each addressing a distinct stage of the indicator assessment workflow:

1. Data Quality and Sensitivity: An automated, rule-based diagnostic framework evaluates the spatial, temporal, and taxonomic completeness of occurrence cubes, flagging potential data limitations before indicator calculation. Complementary group-level cross-validation then quantifies the influence of specific groups, e.g. taxa or time periods, on indicator values.

2. Uncertainty Quantification: To address the frequent omission of variability measures in biodiversity reporting, the package implements a comprehensive bootstrap resampling framework. This allows for the calculation of multiple confidence interval types (percentile, BCa, normal, and basic) tailored to the specific distribution of ecological indicators.

3. Interpretation and Visualisation: The framework links quantitative estimates to an effect classification system that categorizes trends (e.g., increase, stable, decrease). Advanced visualisation techniques, including fan plots and spatial transparency mapping, ensure that uncertainty is explicitly communicated to scientific and policy audiences.

By providing these tools, dubicube ensures that biodiversity indicators are not merely calculated but are evaluated and communicated in a way that supports confident, evidence-based decision-making.