COS 130-1 - Experimentally testing an extinction predictor

Thursday, August 9, 2012: 8:00 AM
D138, Oregon Convention Center
Christopher F. Clements1, N.T. Worsfold2, Phil H. Warren3, Nick Clark3, Ben Collen4, Tim Blackburn5 and Owen L. Petchey6, (1)Animal and Plant Sciences, The University of Sheffield, Sheffield, United Kingdom, (2)University of York, Sheffield, United Kingdom, (3)Department of Animal and Plant Sciences, University of Sheffield, Sheffield, United Kingdom, (4)Zoological Society of London, (5)Genetics, Evolution & Environment, Centre for Biodiversity & Environment Research, London, United Kingdom, (6)1. Department of Evolutionary Ecology and Environmental Studies, University of Zurich, Zurich, Switzerland
Background/Question/Methods

Predictors of time to extinction have been widely used but poorly tested. Optimal Linear Estimation uses a temporal distribution of species sightings to predict when a species will go extinct. The method assumes that frequency of species sightings relates to population size and therefore probability of extinction. It has been suggested that Optimal Linear Estimation provides accurate predictions of extinction time for some species, although it can be susceptible to Type I and Type II error. The use of wild population data to assess the accuracy of a predicted date of extinction is often limited by very approximate estimates of the actual extinction date and limited sighting records. Microcosms provide an experimental setting in which such a method can be trialed against known extinction dates, whilst incorporating a variety of extinction rates created by different temperatures, species identities and species richness’. Here we present the first use of experimental microcosm data to test the accuracy of a sighting based extinction predictor. Long runs of population data were produced from a variety of species and community compositions, preceding eventual extinction caused by competition for resources. Various search regimes (change in the % of the habitat searched through time) were modeled and a series of sighting events generated based on the probability of observing an individual given the area searched and the population density. The predicted date of extinction produced by Optimal Linear Estimation was then compared to the recorded date of extinction and the effect of search effort, search regime, sighting frequency and species identity was quantified.

Results/Conclusions

Our results show that all predictions generated with Optimal Linear Estimation are sensitive both to observer controlled parameters, such as change in search effort, as well as inherent features of the system, such as species identity. Increasing search effort from 1 to 10% rapidly reduces the error associated with predictions, whilst searching more that 10% of a species’ habitat produces negligible increase in predictive precision. Irregular searching produces error much larger than that seen when searching is regular. Combination of low search effort, irregular searching and particular species create error that is orders of magnitude larger than otherwise found. This potential source of error should be considered when using predicted dates of extinction to inform the extinction status of rare species.