AbstractThe widespread uncertainty regarding future changes in climate, socioeconomic conditions, and demographics have increased interest in vulnerability‐based frameworks for long‐term planning of water resources. These frameworks shift the focus from projections of future conditions to the weaknesses of the baseline plans and then to options for reductions in those weaknesses across a wide range of futures. A consistent challenge for vulnerability‐based planning is how to assess the relative likelihood of the occurrence of the multidimensional and codependent uncertainties to which the system or plan is vulnerable. This work proposes a methodological solution to the problem, demonstrated in this case as an extension to Decision Scaling framework. The proposed approach first generates a wide range of futures using stochastic simulators, and then stress tests the system across those futures to identify vulnerabilities relative to stakeholder‐defined performance thresholds. The relative likelihood of the vulnerabilities is then explored using a Bayesian belief network of the knowledge domain of the water resources system. The Bayesian network provides a formal representation of the joint probabilistic behavior of the system conditioned on the uncertain but potentially useful sources of information about the future, including historical trends, expert judgments, and model‐based projections. The proposed approach is applied to compare four design options for a dam project in the Coastal Province of Kenya with respect to the reliability and net present value metrics. Results show that incorporation of belief information helps better distinguishing of the available options, principally by magnifying the differences between the computed net present values.