Please don't misuse the museum: ‘declines’ may be statistical

Evan H Campbell Grant

Detecting declines in populations at broad spatial scales takes enormous effort, and long-term data are often more sparse than is desired for estimating trends, identifying drivers for population changes, framing conservation decisions, or taking management actions. Museum records and historic data can be available at large scales across multiple decades, and are therefore an attractive source of information on the comparative status of populations. However, changes in populations may be real (e.g. in response to environmental covariates) or resulting from variation in our ability to observe the true population response (also possibly related to environmental covariates). This is a (statistical) nuisance in understanding the true status of a population. Evaluating statistical hypotheses alongside more interesting ecological ones is important in the appropriate use of museum data. Two statistical considerations are generally applicable to use of museum records: first without initial random sampling, comparison with contemporary results cannot provide inference to the entire range of a species, and second the availability of only some individuals in a population may respond to environmental changes. Changes in the availability of individuals may reduce the proportion of the population that is present and able to be counted on a given survey event, resulting in an apparent decline even when population size is stable.