Effects of environmental and individual variation on time patterns of extinction in small experimental populations.

Predicting the fate of small populations is essential in ecology, epidemiology and conservation biology. Small populations can go extinct quickly or can develop into large established populations. These can still go extinct through demographic stochasticity, especially when declines in mean population size or large size fluctuations drive them into an extinction vortex. Individual life stage and environmental conditions experienced by founding individuals are two factors expected to determine the fate of small populations. When they don′t go extinct, life stage effects are expected to wash out. Initial environmental differences equally so, unless differences in environmental conditions continue to occur. We used the parthenogenetic Collembola Folsomia candida to investigate such effects in a crossed treatment on 290 replicate populations, each initiated with a single individual. Populations were initiated with founders of three different life stages and subjected to four levels of culling. Their extinction times were recorded in a period of up to 41 weeks after initiation. We fitted the parameters of an age-structured multi-type branching process model to extinction rates observed in the first weeks after initiation. As predicted by the model, extinction probabilities early in the experiment increased with the level of culling. Extinction probabilities were larger in populations founded by non-reproducing individuals. These decreased over time, presumably because of the onset of reproduction. In established populations, extinction probabilities increased with culling level and the effects of the founder stage disappeared, except for a late increase in extinction probability in populations founded by newborns. This increase was not expected and forced us to reconsider when transient effects of the initial state would end. We hypothesize that it is due to effects of size-structured competition in this system. The results show that predictors of extinction probability can be shared between small and established populations. Transient dynamics of extinction risk can be long, with phases even mimicking a stationarity followed by the onset of an extinction vortex.