Refining the stress gradient hypothesis for mixed species groups of African mammals.

Species interactions such as facilitation and predation influence food webs, yet it is unclear how they are mediated by environmental gradients. Here we test the stress gradient hypothesis which predicts that positive species interactions increase with stress. Drawing upon spatially-explicit data of large mammals in an African savanna, we tested how predation risk and primary productivity mediate the occurrence of mixed species groups. Controlling for habitat structure, predation risk by lions and primary productivity affected the frequency of mixed species groups in species-specific ways, likely reflecting distinct stress perceptions. To test whether mixed species groups indicate positive interactions, we conducted network analyses for specific scenarios. Under predation risk, dyadic associations with giraffes were more pronounced and metrics of animal networks changed markedly. However, dyadic association and network metrics were weakly mediated by primary productivity. The composition of mixed species groups was associated with similarities in prey susceptibility but not with similarities in feeding habits of herbivores. Especially predation risk favoured the frequency of mixed species groups and pronounced dyadic associations which dilute predation risk and increase predator detection. While our results provide support for the stress gradient hypothesis, they also highlight that the relative importance of stressors is context-dependent.