Rethinking bias in biodiversity science: sources and solutions

Bias is increasingly a central concern in biodiversity science, yet the term is used inconsistently. A key source of confusion is the failure to distinguish clearly between the dimension in which bias manifests (e.g. spatial or taxonomic) and/or the stage of the observation process that gives rise to it. This distinction matters because the appropriate method for mitigating bias depends primarily on the latter rather than the former. We introduce a theory of observation for biodiversity science that classifies biases according to whether they arise during sampling or subsequent stages of the observation process (e.g. measurement, databasing). This classification is more informative than describing bias by its dimension alone because it directly identifies the mechanisms responsible and therefore the appropriate mitigation strategies. A key consequence of the framework is that an observation process may be a source of sampling or nonsampling bias depending on the scientific question being addressed.