Choosing secure water resource management plans inevitably requires trade-offs between risks (for a variety of stakeholders), costs, and other impacts. We have previously argued that water resources planning should focus upon metrics of risk of water restrictions, accompanied by extensive simulation and scenario-based exploration of uncertainty. However, the results of optimization subject to risk constraints can be sensitive to the specification of tolerable risk, which may not be precisely or consistently defined by different stakeholders. In this paper, we recast the water resources planning problem as a multiobjective optimization problem to identify least cost schemes that satisfy a set of criteria for tolerable risk, where tolerable risk is defined in terms of the frequency of water use restrictions of different levels of severity. Our proposed method links a very large ensemble of climate model projections to a water resource system model and a multiobjective optimization algorithm to identify a Pareto optimal set of water resource management plans across a 25 years planning period. In a case study application to the London water supply system, we identify water resources management plans that, for a given financial cost, maximize performance with respect to one or more probabilistic criteria. This illustrates trade-offs between financial costs of plans and risk, and between risk criteria for four different severities of water use restrictions. Graphical representation of alternative sequences of investments in the Pareto set helps to identify water management options for which there is a robust case for including them in the plan.