Savanna ecosystems are popular subjects for interaction studies. Multiple studies have been done on the impact of elephants on vegetation, the impact of grass and browse availability on animal densities or on competition between herbivore species. Previous studies showed that elephant densities are frequently negatively correlated with densities of tall trees, and that browse and grass availability are correlated with browser and grazer density respectively. Additionally, a competition effect between browse and grass availability has been reported. These relationships are usually analysed by testing direct relationships between e.g., herbivore densities and food availability, without addressing competition effects or other indirect effects. In this study, multiple interactions in a savanna system have been analysed simultaneously using Partial Least Square-Path Modelling (PLS-PM) using mammal and vegetation data from three different wildlife reserves in southern KwaZulu-Natal. The results showed that the processes that three separate models for the three areas provided the best understanding of the importance of the different interactions. These models suggest that elephants had a negative impact on trees, but also on grass availability. The impact is stronger when elephants are not able to migrate during the dry season. Browsers and grazers were correlated with browse and grass availability, but competition between browse and grass was not detected. This study shows that due to the complexity of the interactions in an ecosystem and differences in environmental factors, these interactions are best studied per area. PLS-PM can be a useful tool for estimating direct, indirect, and cascading effects of changing animal densities in conservation areas.