COS 138-3
Estimating extinction risk from presence/absence data with observational uncertainty: Development and evaluation of a flexible modeling framework applied across systems

Friday, August 15, 2014: 8:40 AM
Bataglieri, Sheraton Hotel
Christopher P. Weiss-Lehman, Department of Ecology and Evolutionary Biology, University of Colorado, Boulder, CO
Kendi F. Davies, Department of Ecology and Evolutionary Biology, University of Colorado at Boulder, Boulder, CO
Christopher F. Clements, Animal and Plant Sciences, The University of Sheffield, Sheffield, United Kingdom
Brett A. Melbourne, Department of Ecology and Evolutionary Biology, University of Colorado at Boulder, Boulder, CO
Background/Question/Methods

We are currently undergoing an extinction crisis of such large proportions some are classifying it as a mass extinction on par with only five others in Earth’s history.  Any effort to mitigate this crisis must necessarily begin with accurate assessments of the risk status of species.  Ecologists have many tools to accomplish this, but these methods typically require data on population trends through time, life history traits, and interactions between species. However we lack these data for the majority of species. To address this problem we have developed a model for the analysis of extinction risk from presence/absence data subject to observational uncertainty.  Here we present the development and assessment of the model using systems along a gradient of biological realism.  We first evaluate the model against simulated data across a range of parameter values.  We then apply the model to a data set of protist microcosm communities consisting of four levels of species richness and two different temperature treatments.  Finally, we use the model to estimate extinction risk in beetles from a large-scale fragmentation experiment at Wog Wog, Australia.

Results/Conclusions

The model explicitly incorporates extinction and observation probabilities in a likelihood equation that can be applied in either a frequentist or Bayesian framework and is flexible to system-specific covariates and experimental effects. Using simulated data over a range of parameter combinations, we show the model is robust to varying conditions and can distinguish between temporal variation in pattern and process mechanisms.  Applying our model to the protist data we estimate the effects of temperature and community composition on extinction risk.  In this data set, the time of final extinction for each species in each microcosm is known and the model results recover these dynamics well.  Finally, we apply the model to a beetle data set from Wog Wog, a large-scale habitat fragmentation experiment.  We recover the varying effects of fragment size on different species and demonstrate how to extend the model to a multispecies framework to uncover the interaction between habitat fragmentation and species traits in determining extinction risk.  We conclude that our model is robust even when only using presence/absence data and can be used for many data deficient species to determine those of greatest conservation concern.