Using a model system to test for effects of environmental variance and autocorrelation on population establishment: C. elegans and temperature stochasticity

Background/Question/Methods

Environmental stochasticity has large impacts on the probability that an endangered population will go extinct, or that a newly introduced non-native species will establish and become invasive. While we have a good theoretical grasp of the effects of uncorrelated environmental variance on population dynamics, most environmental conditions, such as temperature, are also strongly autocorrelated in time, even after predictable trends have been removed. The effect of such autocorrelation is much less understood. Previously we developed theory which suggests that small introduced populations subject to autocorrelated environmental variation may be more likely to establish than populations subject to environmental signals with the same variance but lower autocorrelation. We test these predictions using laboratory populations of Caenorhabditis elegans.  This model system offers significant advantages in testing theory that ultimately would be applied to systems for which experimentation is difficult or unethical (i.e., endangered and invasive species). We subjected these populations to time-varying temperature signals where we manipulated the variance and autocorrelation of the series, but held the mean temperature constant. We measured the population size after 4 days of exposure to a temperature signal that varied every 20 minutes. 

Results/Conclusions

We compared the results of these experiments to the predictions of our general model of the effects of environmental stochasticity on population establishment, and to the predictions of a stage-structured model developed specifically for this C. elegans system, where development time, and therefore transition probability, was temperature-dependent. As predicted by classical theory of the effects of uncorrelated environmental variance on population dynamics, we find low population growth and a failure to establish under conditions of a high variance, uncorrelated temperature signal, whereas populations establish easily under constant conditions or low, uncorrelated variance.  As we previously predicted, we find little effect of autocorrelation at low variance: populations increase almost as rapidly as under constant or uncorrelated conditions. In addition, there may be no effect of autocorrelation when signals have large variance. Populations under these conditions fail to increase. However, at intermediate variance, autocorrelation may alter the outcomes as predicted by theory that incorporates only uncorrelated variance.  The exact effect of autocorrelation for signals with intermediate variance depends on the overall mean of the environmental series that affects population growth, as suggested by previous theoretical work.