Accounting for observation reliability when inferring extinction based on sighting records

Given the current increase in global extinction risk of species, methods that are able to estimate the probability and the time of extinction based on a sighting record of a rare or poorly studied species are becoming increasingly important. One of the major obstacles when using such methods is that many sighting records are burdened with uncertain or controversial observations. What is accepted as a valid sighting can have a substantial effect on resulting predictions. Here we present a simple modification to an existing method that allows for the inclusion of specific sighting reliabilities of individual observations into predictions of the likelihood and the time of extinction. The approach is applied to the sighting records of four presumably extinct bird species. Results indicated that the Ivory-billed Woodpecker (Campephilus principalis) and the Eskimo Curlew (Numenius borealis) may be considered as extinct, even when recent controversial sightings are included in the analysis. The Nukupu`u (Hemignathus lucidus) and O`ahu `Alauahio (Paroreomyza maculata) could potentially still be extant, although with a low probability of persistence. The major advantage of the presented method is that, instead of applying some arbitrary threshold for the sighting reliability that is considered acceptable, it recognizes and incorporates the reliability of each observation into the resulting predictions, by estimating the most likely number of observations and the most likely time of the last observation. Its simplicity facilitates its application for assessments of a larger number species or populations, and makes it accessible as a decision tool.