1. Considerable evidence from around the world shows that achieving good ecological outcomes in rivers from programmes of measures in catchments is difficult. There are a number of reasons for this, which we discuss, but here we focus primarily on the question, ‘Is the knowledge base adequate?’
2. We develop further the thesis that catchments and receiving waters are truly complex systems in which there are fundamental limits to knowledge. Our sampling and data analysis practices come with strong biases and inbuilt ‘framing’ assumptions about the nature and values of ecosystem processes that underestimate complexity and uncertainty.
3. If we reframe our assumptions to think of the problem in terms of the properties of complex systems, then we can rethink our attitudes to uncertainty, causal thickets (multiple stressors) and cross-scale effects, and we can begin to develop new definitions of what constitutes a ‘good’ ecological outcome. Dealing with inherent variability in data then becomes less of a problem with controlling ‘noise’ and more of a problem of understanding system dynamics. The presence of adaptive dynamics and self-organisation in complex systems means that uncertainties will always be large and knowledge will be partial and that such systems are fundamentally not computable.
4. We show that small-scale ‘noise’ in ecosystems is an inherent property of non-equilibrium systems with predominant advection, reaction–diffusion dynamics. Flow paths in catchments and the dynamics of receiving waters have fractal properties. Fractal dynamics indicate that multiple, cross-scale interactions are a characteristic of these systems. Small-scale connectivity is an important (and, from a management point of view, underused) aspect of pattern and process in such systems.
5. In an environment of such complexity, models will be flawed and predictions uncertain. It is therefore necessary to develop new indicators of connectivity and ecological complexity so that indicators of system-level progress may be found to assist with an improved process of adaptive management that is trend-orientated as well as outcome-orientated.