Deep uncertainties resulting from climate change, demographic pressures, and rapidly evolving socioeconomic conditions are challenging the way that water planners design and operate large-scale infrastructure systems. Conventionally, water infrastructures have been developed using stationary methods, assuming that the underlying uncertainties can be derived from historical data or experience. However, these methods are less useful under deeply uncertain climate and socioeconomic conditions, in which the future can be substantially different from the past and cannot be expressed by well-defined probability distributions. The recognition of deep uncertainties in long-term water resources planning has led to the development of “decision-analytical” frameworks that do not require predictions or prior probabilistic inference about the future. Instead, these approaches seek for alternatives that perform well across a broad range of conditions (robust) and can adapt to changing conditions (flexible). This dissertation aims to develop three new decision-analytical frameworks that build upon the previous work. The first study presents a generalized framework for water infrastructure design under climate change using regret-based robustness criterion and compares the findings to more conventional, predict-then-act based analyses of infrastructure design. The method is demonstrated for the design of a run-of-the-river hydropower system in Malawi. The second study further develops the framework by considering multiple climatic, demographic, and socioeconomic uncertainties in the context of a water supply design project in the Coastal Kenya. This improved framework incorporates a Bayesian belief network to blend multiple sources of subjective information from model projections and expert opinions elicited from stakeholder workshops. The third framework develops a decision-analytical approach for flexible river basin planning under climate change and applies to the problem of long-term water supply and irrigation planning in the Niger River Basin. The framework makes use of a stochastic programming model to search for optimal planning pathways under a wide range of scenarios that represent both natural climate variability and climate changes. In this process, the framework explores uncertain beliefs associated with the probability weights assigned to each scenario and identifies “belief dominant” pathways that are insensitive to underlying probabilistic assumptions and are more promising based on climate projections.