Anthropogenic disturbance and rainfall variation threaten the stability of plant-ant interactions in the Brazilian Caatinga

Climate change is projected to exacerbate the effects of anthropogenic disturbance, with negative impacts on ecosystem stability and functioning. We evaluate the additive and combined effects of chronic anthropogenic disturbance (CAD) and rainfall variation on the temporal stability of mutualistic EFN-bearing plant-ant networks in a Caatinga dry forest. We evaluated whether changes in the stability of these interactions are driven by changes in the stability of the communities of partners involved and/or in ant behavior. We sampled EFN-bearing plant-ant networks in sixteen 20 x 20 m plots distributed across CAD and rainfall gradients. The stability of EFN-bearing plant and attendant-ant communities were measured as the inverse of temporal differences in their community structure and composition. We also computed the stability of EFN-bearing plant-ant networks by measuring the inverse of temporal differences in network specialization metrics. We found that, in general, the structure and composition of plant and ant interacting communities were similarly stable along both environmental gradients. Only CAD and its interaction with rainfall affected the temporal stability of EFN-bearing plant diversity, which declined as CAD increased, with a more pronounced relationship in wetter areas. However, variation in levels of CAD and, to a lesser extent, rainfall greatly modulated the stability of EFN-bearing plant-ant network specialization. CAD reduced the stability of network generality (specialization at the ant level), an effect that was much stronger in wetter areas. Meanwhile, the stability in network vulnerability (specialization at the plant level) decreased with the increase of CAD and the decrease of rainfall levels. Finally, there was a trend of decreasing stability in specialization of the overall network with increasing CAD. Our results suggest that changes in the structure of interaction networks are mainly driven by a switch in ant behavior rather than by changes in the structure and composition of plant and ant communities between years.