Rivers are complex systems for which it is hard to make reliable assessments of causes and responses to impairments. We present a holistic risk-based framework for river ecosystem assessment integrating all potential intervening processes and functions. Risk approaches allow us to deal with uncertainty both in the construction of indicators for magnitude of stressors and in the inference of environmental processes and their impairment. Yet, here we go further than simply replacing uncertainty by a risk factor. We introduce a more accurate and rigorous notion of risk with a transcription of uncertainty in causal relationships in probability distributions for the magnitude of impairment and the weight of different descriptors, with an associated confidence in the diagnostic. We discuss how Bayesian belief networks and Bayesian hierarchical inference allow us to deal with this risk concept to predict impairments and potential recovery of river ecosystems. We developed a comprehensive approach for river ecosystem assessment, which offers an appealing tool to facilitate diagnosis of the likely causes of impairment and predict future conditions. The ability of the risk approaches to integrate multi-scale quantitative and qualitative descriptors in the identification of multiple stressor sources and pathways in the stream network, and their impairment of specific processes and structures is illustrated for the national-level risk analysis for hydromorphology and pesticide pollution. Not only does the risk-based framework provide a more complete picture of environmental impairments, but it also offers a comprehensive, user-friendly tool to instruct the decision process.