Using a Bayesian network model to assess ecological responses to hydrological factor interactions

Abstract

Interactions between environmental factors can dramatically influence the relationships between a species and its environment. However, multiple types of interaction are possible, and as such, some may be overlooked. We used a Bayesian network to model the response of a riparian tree species, Eucalyptus camaldulensis Dehnh., to the interactive influences of hydrological factors. We used a novel conceptual framework, which included not only synergistic and antagonistic interactions but also qualitative interactions (i.e. the effect of an environmental factor that switches from positive to negative or vice versa depending on the value of another factor). Synergistic, antagonistic and qualitative interactions were all detected in the response of E. camaldulensis to hydrological factor interactions. The predicted influence of environmental factors varied dramatically when interactions were considered. In some instances, the likelihood of a certain outcome differed by greater than 80% when interactions were considered. For example, the negative impact of grazing on E. camaldulensis was far greater in areas of deep groundwater relative to areas of shallow groundwater. Interactions also led to qualitatively different predictions (i.e. a qualitative interaction occurred). For example, frequent wetland inundation had a positive influence on tree vigour in wetlands minimally impacted by weirs, but a negative influence in wetlands close to weirs. Considering interactions may therefore substantially change understanding of ecohydrological relationships. Thinking of interactions between factors as potentially qualitative, and not only as synergistic and antagonistic, is likely to be important in situations where multiple management interventions are proposed.