An experimental test of a Bayesian method for inferring extinction with varying search efforts
Determining whether a species is extinct or extant is notoriously difficult, but is fundamental to both our understanding of biodiversity loss, and our ability to implement effective conservation measures. Many methods have been proposed in an attempt to infer quantitatively whether a species has gone extinct, with many seeking to do so by using sets of historic sighting events. Until recently, however, no methods have been proposed that explicitly take into account search effort (the proportion of a habitat searched when looking for a species), a key determinant of if/when historic sighting events have occurred. Here we present the first test of a recently proposed Bayesian approach for inferring the extinction status of a species from a set of historic sighting events where the search effort that has produced the sightings can be explicitly included in the calculation. We utilize data from a highly tractable experimental system, as well as simulated data, to test whether the method is robust to changing search efforts, and different levels of detectability of a species. We find that, whilst in general the method performs well, it is susceptible to both changes in search effort through time, as well as how detectable a species is. In addition, we show that the value of the prior expectation that the species is extant has a large impact on the accuracy of the methods, and that selecting correct priors is critical for accurate inference of extinction status.