Biotic interactions and ecological predictability

Ecological systems are exceedingly difficult to predict. They integrate many different species, their variations in abundances, their interactions, and the influences from external factors like climate or human disturbances. There exist terrestrial, marine, riverine ecosystems with very different species that interact in different ways between them and with their environment. Understanding and predicting their behaviour, despite the challenge it poses, is fundamental to developing biological conservation and restoration programs in the face of global change. How do we move ecology towards a more predictive science? In particular, how do we robustly anticipate which species in a community will increase in numbers, or which ones will tend to go extinct? Changes in species abundances are modulated by their fertility and mortality rates, as well as by the interactions with other species - who eats whom, who pollinates the flowers of which plant, etc. Ecologists know much about which species interact with each other, and why. But how these interactions translate to variations in species abundances is much less clear. It is common to simplify such effects by assuming that a certain interaction always has a constant effect in the species involved. But in reality this is likely to vary - for example, it is not the same for a predator, say a wolf, to consume a young deer, who still cannot reproduce, than to consume a female who is about to get pregnant or give birth. Understanding and quantifying the variation in the effects of biotic interactions is a key step towards predicting changes in species abundances. The HOROS project is named after the ancient Greek word for boundary or limit. With it, we want to move closer to the limits of predictability in ecological sciences: what can we predict, to what extent, and with what level of confidence. HOROS will be the first project to explicitly quantify how different types of interactions (predation, competition, mutualism) vary in their demographic effects across different species and ecosystem types. We will do so by developing statistical time-series methods that we will apply to data sets from terrestrial, marine, and riverine ecosystems. The analysis of how interactions vary in their demographic effects will be the first step towards building predictive models that account for the full uncertainty of ecological processes. More variable interactions will likely imply that ecological systems are harder to predict robustly, or that the time window that can be accurately predicted is shorter. We will build and validate predictive models using our interaction data, to understand which ecosystems can be predicted with higher confidence, and which methodologies would be more appropriate for the task. Overall, we expect to contribute to developing more robust and realistic predictions of ecological systems, which we urgently need in the face of the current climate and biodiversity crisis.