Bias assessments to expand research harnessing biological collections

Biological collections are arguably the most important resources for investigations into the impacts of human activities on biodiversity. However, the apparent opportunities presented by museum-derived datasets have not resulted in consistent or widespread use of specimens in ecology outside phenological research and species distribution modeling. We attribute this gap between opportunity and application to biases introduced by collectors, curators, and preservation practices and an imperfect understanding of these biases and how to mitigate them. To facilitate broader use of specimen-based data, we characterize collection biases across key axes and explore interactions among them. We then present a framework for determining the bias assessments needed when extracting data from biological collections. We show that bias assessments required by particular ecological studies will depend on the response variables being measured and the predictor axes of interest. We argue that quantification of biases in specimen-derived datasets is needed to facilitate the widespread application of these data.