Most of the world’s ecosystems suffer from stress caused by human activities such as habitat destruction, fragmentation, overexploitation of species and climate change. These factors affect the reproduction and/or survival of individual species as well as interactions between species in ecological communities. Forthcoming effects of this are altered abundances, direct species loss, and indirect cascading extinctions, with yet largely unknown consequences on community structure and functioning. Today, biodiversity loss is of global concern since human society and welfare depend upon resources and services provided by ecosystems. The importance of considering entire ecological communities as a target for conservation and management has been increasingly recognized due to the interdependencie  of species. Our ability to make predictions of the response of ecological communities to stress and biodiversity loss is in need of a deeper understanding of how structure and dynamical processes contributes to the functioning and stability of a community. In this thesis I use mathematical theory and dynamical models to study the response of community structure and resilience to a variety of disturbances affecting species and species interactions, ranging from small perturbations (Papers I-II) to large perturbations (species extinctions, Papers IIIIV).

In Paper I we develop Community Sensitivity Analysis (CSA) as an analytical tool to study how a small permanent perturbation to the intrinsic growth rate, or mortality rate, of species is expected to affect i) the resilience (return rate) and ii) the structure (distribution of species equilibrium abundances) of an ecological community. Species interactions are described using Lotka-Volterra predator-prey dynamics. We apply CSA on the pelagic food webs of Lake Vättern and the Baltic Sea, respectively, and find that a change in the mortality rate of large-bodied species has a higher impact on community resilience and structure, compared to a perturbation to small-bodied species. However, analyzing the effect of a proportional change to the growth or mortality rate of species (elasticity analysis) shows that smallbodied species have proportionally larger effects on species equilibrium abundances, but not on resilience. CSA can also be used to study the effect of permanent (absolute or proportional) changes to inter- and intraspecific interaction strengths. For the two pelagic systems used in this study, CSA reveal that changes in the effect of a prey on its consumer tend to affect community structure and resilience significantly more than changes in the effect of a predator on its prey.

In Paper II we assess the importance of rare species for the structure and resilience of ecological communities. First we show analytically, for a two species predator-prey system, that a change in the intrinsic growth rate of the rare species affect resilience more than a change in the growth rate of the common species. To test the generality of these results we next apply CSA on complex model food webs. In the analysis we distinguish between four trophic groups, each including only species with a similar trophic position, to separate the effect of abundance from the trophic position of species. Using mixed effect models we find support for our analytical predictions. More precisely, we find a strong negative relationship between the importance (sensitivity) of a species and its equilibrium abundance within all consumer groups and a weaker, but significant, relationship for producer species. The results from this study suggest that rare species can act as keystones through their effect on both community resilience and community structure, regardless of its trophic position.

In Paper III we evaluate the risk of food web collapse caused by different trait-based extinction scenarios. In previous studies, groups of species, e.g. rare species, large-bodied species and top predators, have been identified to be relatively more prone to extinctions and other studies have found that extinctions of such species have comparably small effects on the remaining community. Using mathematical models of species dynamics we study the response of ecological communities to species removal (i.e. the proportion of species needed to be primarily removed to cause a 50% reduction in species richness, R⁵⁰) when species are sequentially removed from the food web based on eight different traits. We show, contrary to some previous studies of sequential extinction simulations, that communities can be very vulnerable to realistic species loss. We furthermore find that the response of communities seems to depend on whether the extinction sequence follows a bottom-up or top-down direction, making it difficult to identify one particular extinction sequence as the most important/severe sequence.

Finally, in Paper IV we aim to identify traits of species that can be used to identify keystone species, in terms of causing the highest proportion of secondary extinctions following their loss, in food webs with different degree of disassembly. Moreover, we analyze if the loss of a species that triggers a cascade of many secondary extinctions are the same species being identified as a keystones using Community Sensitivity Analysis. To answer these questions we randomly remove species from a set of 100 model communities. We analyze the relationship between the number of secondary extinctions following the randomly removed species and a range of species traits in communities where i) 75-100% of the initial number of species remain, ii) 50-75% of all species remain, iii) 25-50% of all species remain and iv) only 0-25% of all species remain. We find that the variation in secondary extinctions explained using species traits increases when the degree of food web disassembly and food web connectance are taken into account. The most important trait varies for different degrees of food web disassembly and also depends on whether basal species can go primarily extinct or not. However, due to correlation between most important traits, we conclude that the key status of different traits is rather robust against structural changes in the model food webs. Interestingly, food webs seem to be most sensitive to a random species loss after the loss of more than 25% of all initial species, suggesting that there is a threshold from which secondary extinctions increases. We also conclude that species being identified as keystones, based on the effect of their loss, are to some extent the same species being identified as having the largest effect on community structure and resilience, respectively, following a small perturbation.