By Claire Asher, on 5 January 2015
Classifying a species as either extinct or extant is important if we are to quantify and monitor current rates of biodiversity loss, but it is rare that a biologist is handy to actually observe an extinction event. Finding the last member of a species is difficult, if not impossible, so extinction classifications are usually estimates based on the last recorded sightings of a species. Estimates always come with some inaccuracy, however, and recent research by GEE academics Dr Ben Collen and Professor Tim Blackburn aimed to investigate how accurate our best estimates of extinction really are.
Using data from experimental populations of the single-celled protist Loxocephalus, as well as wild populations of seven species of mammal, bird and amphibian, the authors tested six alternative estimation techniques to calculation the actual date of extinction. In particular, they were interested in whether the accuracy of these estimates is influenced by the rate of population decline, the search effort put in to find remaining individuals and the total number of sightings of the species. The dataset included very rapid declines (40% a year in the Common Mist frog) and much slower ones (16% per year in the Corncrake), and different sampling regimes.
Their results showed that the speed of decline was a crucial factor affecting the accuracy of extinction estimates – for experimental laboratory populations, estimates were most accurate for rapid population declines, however slow population declines in wild populations tended to produce more accurate results. The sampling regime was also important, with larger inaccuracies occurring when sampling effort decreased over time. This is probably a common situation for many species – close monitoring is common for species of high conservation priority, but interest may decrease as the species becomes closer and closer to extinction. The total number of sightings was also an important factor – a larger number of sightings overall tended to produce more accurate estimates.
Finally, the estimation technique also influenced accuracy, but only in interaction with the other variables mentioned above. Some methods fared best for rapid population declines, others for slower ones. Many of the methods fare poorly when sampling effort changes over time, particularly if it decreases, although they were relatively robust to sporadic, opportunistic sampling regimes. Overall, optimal linear estimation, a statistical method which makes fewer assumptions about the exact pattern of sightings, produced the most accurate results in cases where more than 10 sightings were recorded in total.
This study highlights the challenges faced by ecologists trying to determine whether a species has gone extinct or not. Sightings of rare species are often opportunistic, and only rarely are they part of a systematic, long-term monitoring program. Thus, methods that produce accurate results in the face of changing or sporadic search efforts are of key importance to conservationists. If the history of a species’ population declines and of the sampling effort are known, then statistical estimates can be selected which provide the best estimates for the particular situation. However, this information is rarely available and so using techniques that can provide accurate estimates for a range of different historical scenarios are likely to be of most use in predicting extinction status. Ultimately, it is extremely useful for conservationists to know whether a species is extinct or not, but estimates will always be subject to error except in rare cases (such as the passenger pigeon, for example) where the extinction event is observed first hand. There will always be cases of species turning up years after they were declared extinct, and no estimate will ever be perfect, but understanding the sources of error and the best methods to use to minimise it can be of great benefit in reducing the frequency with which that happens.
This research was made possible by funding from the Natural Environment Research Council (NERC).