Modeling critical transitions in natural systems: Can extinctions be predicted?

Understanding population dynamics in response to slowly changing environmental drivers such as climate and land use changes is essential for conservation efforts and for understanding the resilience of an ecosystem. Current efforts have focused on anticipating thresholds in populations that result in transitions to alternative dynamical states. This can be done by fitting complex statistical models to the time series data. An alternative approach is to monitor statistical indicators, but this method is not well tested. We used both statistical models and indicators to determine if extinction thresholds were present and detectable in environmental data sets. We applied these methods to both paleolimnological data of a known local extinction of a dominant algal species (Cymbella cymbiformis) driven by climate change and monitoring data of a declining population of game bird, the bobwhite quail (Colinus virginianus) that is experiencing habitat deterioration due to land use change. In both cases, we found that a common extinction threshold was present; the crossing of the threshold could have been predicted by monitoring suitable statistical indicators and with time series analysis. Both the model and the statistical indicators correctly predicted that C. cymbiformis was committed to a local extinction evidenced in the paleorecord and the model suggests that C. virginianus crossed the extinction threshold in 2011(2003–2032, 95% CI) and will undergo a local extinction in the near future. Our results show that both methods are useful in determining extinction thresholds despite drastically differing taxa, drivers of extinction, and time frames being evaluated.