Climate oscillation impacts on water supply augmentation planning.

Climate oscillations ranging from years to decades drive precipitation variability in many river basins globally. As a result, many regions will require new water infrastructure investments to maintain reliable water supply. However, current adaptation approaches focus on long-term trends, preparing for average climate conditions at mid- or end-of-century. The impact of climate oscillations, which bring prolonged and variable but temporary dry periods, on water supply augmentation needs is unknown. Current approaches for theory development in nature-society systems are limited in their ability to realistically capture the impacts of climate oscillations on water supply. Here, we develop an approach to build middle-range theory on how common climate oscillations affect low-cost, reliable water supply augmentation strategies. We extract contrasting climate oscillation patterns across sub-Saharan Africa and study their impacts on a generic water supply system. Our approach integrates climate model projections, nonstationary signal processing, stochastic weather generation, and reinforcement learning-based advances in stochastic dynamic control. We find that longer climate oscillations often require greater water supply augmentation capacity but benefit more from dynamic approaches. Therefore, in settings with the adaptive capacity to revisit planning decisions frequently, longer climate oscillations do not require greater capacity. By building theory on the relationship between climate oscillations and least-cost reliable water supply augmentation, our findings can help planners target scarce resources and guide water technology and policy innovation. This approach can be used to support climate adaptation planning across large spatial scales in sectors impacted by climate variability.

Neuro-Evolutionary Direct Policy Search for Multiobjective Optimal Control.

Direct policy search (DPS) is emerging as one of the most effective and widely applied reinforcement learning (RL) methods to design optimal control policies for multiobjective Markov decision processes (MOMDPs). Traditionally, DPS defines the control policy within a preselected functional class and searches its optimal parameterization with respect to a given set of objectives. The functional class should be tailored to the problem at hand and its selection is crucial, as it determines the search space within which solutions can be found. In MOMDPs problems, a different objective tradeoff determines a different fitness landscape, requiring a tradeoff-dynamic functional class selection. Yet, in state-of-the-art applications, the policy class is generally selected a priori and kept constant across the multidimensional objective space. In this work, we present a novel policy search routine called neuro-evolutionary multiobjective DPS (NEMODPS), which extends the DPS problem formulation to conjunctively search the policy functional class and its parameterization in a hyperspace containing policy architectures and coefficients. NEMODPS begins with a population of minimally structured approximating networks and progressively builds more sophisticated architectures by topological and parametrical mutation and crossover, and selection of the fittest individuals concerning multiple objectives. We tested NEMODPS for the problem of designing the control policy of a multipurpose water system. Numerical results show that the tradeoff-dynamic structural and parametrical policy search of NEMODPS is consistent across multiple runs, and outperforms the solutions designed via traditional DPS with predefined policy topologies.

When timing matters-misdesigned dam filling impacts hydropower sustainability.

Decades of sustainable dam planning efforts have focused on containing dam impacts in regime conditions, when the dam is fully filled and operational, overlooking potential disputes raised by the filling phase. Here, we argue that filling timing and operations can catalyze most of the conflicts associated with a dam's lifetime, which can be mitigated by adaptive solutions that respond to medium-to-long term hydroclimatic fluctuations. Our retrospective analysis of the contested recent filling of Gibe III in the Omo-Turkana basin provides quantitative evidence of the benefits generated by adaptive filling strategies, attaining levels of hydropower production comparable with the historical ones while curtailing the negative impacts to downstream users. Our results can inform a more sustainable filling of the new megadam currently under construction downstream of Gibe III, and are generalizable to the almost 500 planned dams worldwide in regions influenced by climate feedbacks, thus representing a significant scope to reduce the societal and environmental impacts of a large number of new hydropower reservoirs.