Better safe than sorry

common to all species (although the evidence for this appears weak; Flather et al. 2011), then the SAFE index is merely a measure of N. The need for the logarithmic transformation in the SAFE index is unsupported. Clements et al. claim that the SAFE index improves communication of “distance” from extinction. However, the general public, donors, and policy makers will understand the implications of N, and how it departs from MVP, much more easily than those of log(N/MVP). Clements et al.’s result that the SAFE index correlates more strongly with IUCN classes than with historical range reduction is unsurprising. Population size, and hence the SAFE index, must correlate with IUCN classes because population size is one of the major criteria used in the IUCN classification. In contrast, historical range reduction will only correlate with IUCN classifications as far as it correlates with current reductions in range or population size, or with some other measure that is used in the IUCN criteria. A further concern is that the analysis was based on linear models. Apparent low correspondence for an index could arise from unmodeled non-linear relationships with IUCN classes. Clements et al. also claim that “the SAFE index provides a more meaningful and fine-grained interpretation of the relative threat of species extinction than do the IUCN threat categories alone”. Although IUCN classes are correlated with actual extinctions (Keith et al. 2004), Clements et al. provide no evidence that the SAFE index is “more meaningful” than assessments based on IUCN criteria because, for example, they did not analyze independent extinctions. The acronym for the index (“species’ ability to forestall extinction”) implies that it measures whether it is worth trying to save a species. This implication is stated explicitly by Clements et al.: “Practitioners of conservation triage may want to prioritize resources on the Sumatran rhinoceros (Dicerorhinus sumatrensis) instead of the Javan rhinoceros (Rhinoceros sondaicus)” based on differences in their SAFE index values. This statement is not supported by the published paper, which focused on a measure of threat. Optimal prioritization depends on the management objective, the efficiency with which spending resources on different species meets that objective, the uncertainties in the measures of efficiency, and the available budget (McCarthy et al. 2008, 2010). For example, aiming to minimize the number of extinct species can lead to very different allocations from those obtained when aiming to minimize the number of threatened species, even in the absence of any other differences (McCarthy et al. 2008). In conclusion, we argue that conservation prioritization based solely on population size is unsubstantiated and potentially misleading. Michael A McCarthy, Georgia E Garrard, Alana L Moore, Kirsten M Parris, Tracey J Regan, and Gerard E Ryan School of Botany, The University of Melbourne, Parkville, Australia (mamcca@unimelb.edu.au); WWF Cambodia, Phnom Penh, Cambodia