Toward Joint Information Selection and Policy Learning in Water Resources Management

The water resources literature has made major advances in computational tools for infrastructure management, developing increasingly sophisticated adaptive and many-objective control frameworks that improve how decisions are optimized under uncertainty and change. Yet most water infrastructure operating policies still condition actions on narrow information sets, typically reservoir storage and season, and more rarely, a streamflow forecast. Meanwhile, monitoring networks, forecasting systems, and high-resolution models now generate far richer signals about current conditions and future disturbances. We argue that the choice of what information conditions policy decisions, that is, its information representation, remains an underexplored but fundamental design step of planning and operating policies. Because policies map observed information to actions, their effectiveness is ultimately bounded by the quality and relevance of that representation. This review frames information selection as a core problem in water decision-making and synthesizes methods to select or learn policy inputs. Among the key contributions of this paper, we propose a classification of representative prior studies as “a priori”, “a posteriori”, and “joint” information selection approaches. We review consensus findings and highlight open challenges, focusing on settings in which system performance is particularly sensitive to information selection, for example, during extremes, under nonstationarity, and in multi-objective contexts. We further extend the discussion to long-term planning, where the problem of selecting the information representation is complicated by deep uncertainty in climate and socio-economic futures. We conclude by offering a pathway to integrate information selection lessons when designing monitoring programs and models to support adaptive decisions in a changing world.

Duke Scholars

Author Marta Zaniolo Civil and Environmental Engineering