Palaeoecology offers a potentially powerful approach to both understand ecosystem dynamics and the effects of multiple anthropogenic stressors on wetland ecosystems because of its capacity to extend temporal records of ecosystem condition back in time for centuries and even millennia. However, attempts to use palaeo-approaches to derive long-term ecological records and make general inferences about ecological processes are often been hampered by focus on a single site, or a limited number of sites, which means that records incorporate idiosyncratic site-specific responses that cannot be used to infer representative responses to broad-scale drivers or generalizable ecosystem dynamics due to a lack of replication. Conversely, attempts to synthesize records from multiple studies at a regional scale are often hampered inconsistent methods, uncertain chronologies and spatial variation in the underlying character of the studied systems and in the intensity and nature of the stressors that affect them. This study reports on the development of a Bayesian Belief Network (BBN) to predict the impacts of catchment disturbance and flow regulation on floodplain wetlands. The approach conceptualizes wetland ecosystem response as dependent on underlying physical character (depth, hydroperiod) and on the nature and intensity of the proximal stressors arising from distal drivers (principally catchment disturbance and river regulation) that combine to from a dynamic ‘driver surface’. The nature of this relationship is then tested and refined by the development of a BBN and comparing the outcome with changes observed in palaeoecological records from billabongs of the floodplains on rivers in the southern MDB.