It is obvious that rarity plays an important role in the extinction risk of species. It is less frequently acknowledged that the relationship between rarity and extinction risk is largely an expression of random chance. This contribution of random chance to species extinction and extirpation probability is sometimes ignored, often in favor of deterministic, trait-based explanations for extinctions. To address the influence of stochasticity on species extinction risk, we applied a null model approach to analyze local macroinvertebrate extinctions in a series of experimental terrestrial ecosystems that were subject to experimental habitat destruction. Each mesocosm included sixteen Solidago altissima individuals serving as habitat for a regional total of 110 arthropod taxa. Mesocosms were surveyed for macroinvertebrate diversity prior to habitat destruction, with additional surveys occurring after the habitat loss. Observed macroinvertebrate extirpations were compared to null-expected extinctions formulated through a sample-based rarefaction model. Under the assumptions of the model, species loss is dictated solely by the random removal of individual plants from each mesocosm. As a result, observed extinctions that deviate from our random expectations may be evidence of deterministic and trait-based extinction processes.
Results/Conclusions
Our initial results indicate that while we observed greater than expected losses of taxa, a majority of the observed extinctions were consistent with stochastic predictions based on the random loss of habitat. We compared null-expected extinctions to observed extinctions and found that null-expected species richness is a weak predictor of observed richness immediately after disturbance (R2 = 0.34, p = 0.08), but is a significant predictor three weeks after disturbance (R2 = 0.61, p = 0.007), suggesting an important role for species abundance in determining local extinction risk. Interestingly, we observed significant losses of biodiversity at the regional (across-site) level despite most observed extirpations being consistent with random chance; species richness deviated significantly further from null-expectations at the regional level than the local level. Ultimately, our findings suggest that abundance should be considered in assessing a species extinction risk. While many observed extinctions were stochastic, the relative dominance of these processes are scale-dependent and several abundant species experienced extinction, suggesting an important role for trait-based determinism. In sum, our results suggest an important role for null models in assessing extinction risk when information is limited, although we find a surprisingly important role for species traits in a simulated random disturbance.