Food Webs, Models and Species Extinctions in a Stochastic Environment

In light of the current global mass extinction of species, ecologists are facing great challenges. In order to reverse the path towards additional extinctions early warning systems to guide management actions need to be developed. However, considering the countless species to monitor and the complexity of interactions affecting species abundances in ecological communities, this is not an easy task. Before this goal can be reached our understanding of how community structure and species interactions interact and affect the risk of extinction of single species needs to be increased. Thus the primary aim of the present thesis is to study this interaction and contribute to a theoretical basis for the identification of extinction prone species.

In paper II it is concluded that spectral analysis of population time series may function as a tool to predict extinctions at an early stage. More specifically, I show that extinction risk of producer species in food webs under influence of uncorrelated environmental stochasticity increases with intensified red-shift of population time series. However, this relationship is strictly context-dependent, which means that a producer with red dynamics might survive in one type of food web, but the same producer species with a similar magnitude of spectral redness can go extinct in another food web where the interactions with other species are arranged in a different manner.

Then I turn to look at which species might be more prone to become endangered or to go extinct in food webs experiencing various types of uncorrelated environmental stochasticity. In paper I I show that producer species are more likely to reach endangered population levels (according to The World Conservation Union, IUCN, criterion), whereas paper III demonstrates that consumer species more frequently go extinct. This seemingly contradiction may be explained by characteristics inherent to many producer species (e.g. high growth rate, short generation time) that enable them to recover from low population levels and thus escape extinction. Furthermore, in both the second and the third paper I show that the structure of food webs as well as the presence, position and direction of a strong interaction between two species in a food web play significant roles in the likelihood of a species reaching endangered population levels or going extinct.

In paper IV I show that small and condensed food webs are likely to express fundamentally different dynamics compared to large and well-resolved versions of the same natural food webs. Starting from a well-resolved version of a real food web, local dynamics of the ecological system change in a non-linear manner, during gradual lumping of the functionally most similar species into aggregated species (or trophospecies), Here it is also suggested that functional redundancy exists in natural food webs. This may imply support for the ?insurance hypothesis? since sequential extinction of one of the species in the functionally most similar pair of species initially did not generate any significant changes in local dynamics of the system.

To sum up, in this thesis I present a prototype of a predictive tool to discover species at risk of going extinct. I also present directions to which type of species to look for and what type of structures and interactions to pay attention to when searching for presumptive victims of extinction in ecological systems. However, the features of the ecological models I have used for my research are in many cases incomplete. For example, my food webs contain relatively few species without competitive interactions subjected to only uncorrelated environmental variability. Further research will have to test the generality of the results and the robustness of the conclusions drawn from them.