A test to know whether species with few records or recent last-sighting dates are likely to be extinct

Inferring Extinction of Mammals from Sighting Records, Threats, and Biological Traits

DIANA O. FISHER and SIMON P. BLOMBERG

Conservation Biology, Volume 26, Issue 1, pages 57–67, February 2012

Quantitative techniques exist for species with five or more sightings, to test whether a species is extinct on the basis of distribution of sightings. However, 70% of purportedly extinct mammals are known from fewer than five sightings. Such models do not include important indicators of the likelihood of extinction like threats, biological traits, search effort, and demography.

The researchers here developed a quantitative method that they based on species’ traits in which they used Cox proportional hazards regression to calculate the probability of rediscovery of species regarded as extinct. They used two versions of the Cox regression model to determine the probability of extinction in purportedly extinct mammals and compared the results of these two models with those of stationary Poisson, nonparametric, and Weibull sighting-distribution models. 

For mammals with five or more sightings, the stationary Poisson model categorized all but two critically endangered (flagged as possibly extinct) species in their data set as extinct. The results with this model were consistent with current categories of IUCN. The scores of probability of rediscovery for individual species in one version of their Cox regression model were correlated with scores assigned by the stationary Poisson model. They used this Cox regression model to determine the probability of extinction of mammals with sparse records.

On the basis of the Cox regression model, the most likely mammals to be rediscovered were the Montane monkey-faced bat (Pteralopex pulchra), Armenian myotis (Myotis hajastanicus), Alcorn's pocket gopher (Pappogeomys alcorni), and Wimmer's shrew (Crocidura wimmeri). The Cox model categorized two species that have recently disappeared as extinct: the baiji (Lipotes vexillifer) and the Christmas Island pipistrelle (Pipistrellus murrayi).

The researchers affirm that their new method can be used to test whether species with few records or recent last-sighting dates are likely to be extinct.