Abstract
Interactions are key drivers of the functioning and fate of plant communities. A traditional way to measure them is to use pairwise experiments, but such experiments do not scale up to species-rich communities. For those, using association networks based on spatial patterns may provide a more realistic approach. While this method has been successful in abiotically-stressed environments (alpine and arid ecosystems), it is unclear how well it generalizes to other types of environments. We help fill this knowledge gap by documenting how the structure of plant communities changes in a Mediterranean dry grassland grazed by sheep using plant spatial association networks. We investigated how the structure of these networks changed with grazing intensity to show the effect of biotic disturbance on community structure. We found that these grazed grassland communities were mostly dominated by negative associations, suggesting a dominance of interference over facilitation regardless of the disturbance level. The topology of the networks revealed that the number of associations were not evenly-distributed across species, but rather that a small subset of species established most negative associations under low grazing conditions. All these aspects of spatial organization vanished under high level of grazing as association networks became more similar to null expectations. Our study shows that grazed herbaceous plant communities display a highly nonrandom organization that responds strongly to disturbance and can be measured through association networks. This approach thus appears insightful to test general hypotheses about plant communities, and in particular understand how anthropogenic perturbations affect the organization of ecological communities.