Abstract
Ecological systems are not simple but composed of many different elements(species, for example) interacting with each other. These networks ofinteractions exhibit structural properties that determine ecological systems’ability to absorb and recover from perturbations. Mappinginteractions along with their changes in time and space is therefore key tounderstand and predict empirical communities' response to global changes.In this thesis, we used plant communities as model systems (i) to explore howspatial patterns may help identify feedbacks loops which make communities morefragile to upcoming changes and (ii) to map species interactions in empiricalcommunities and describe how they change along stress gradients and recover fromperturbations. To do so, we used two datasets documenting plant communities insubalpine meadows (USA) and Mediterranean grasslands (France).Our results show that feedback loops can be inferred to some extent from thespatial patterns of plant communities and hence help identify communities thatmay respond more abruptly to perturbations. Going to a more detailed level ofdescription, plant-plant interactions (as measured through spatial associations)were shown to respond strongly and consistently to stress but exhibited a weakresilience to disturbances.This work shows that plant-plant interactions -- which are linked to the response of the community to perturbations -- can be uncovered using spatial patterns. It paves the way towards a better understanding and a better anticipation capacity of how ecological communities might reorganize when subject to disturbances.