It is widely accepted that water resource management demands an integrated assessment of resource use options, including local and regional impacts on the environment and stakeholders. Multiple issues, stakeholders and scales of system behaviour must be considered, as well as the key disciplines within and between the human and natural sciences. Modelling is a critical tool in integrated assessment. It enables effects of policy interventions, climate forcing and demographics to be predicted (although with some uncertainty), and provides a means of expanding understanding of river basin behaviour. It also acts as a vehicle for social learning among various interest groups. This paper discusses the various frameworks and methods being used for integrated modelling, and their suitability and unfulfilled potential for these purposes. The frameworks include coupled component models, systems dynamics models, metamodels, risk-assessment approaches, Bayesian decision networks, agent-based methods, expert systems and other heuristic knowledge-based techniques. Specific software platforms are not considered but the lessons from software development and implementation are clearly spelt out. The paper presents three Australian case studies in integrated assessment. They vary in their range of catchment/watershed sizes, hydroclimatology, issues of concern and stakeholders engaged. Two of them utilise a coupled component modelling framework and the third a Bayesian decision network approach. The paper illustrates the value, problems and lessons of integrated assessment and modelling. In particular it proposes some Water Resour Manage (2007) 21:351-373 ways to address the challenges of assessing options to obtain more sustainable basin-wide outcomes.