Using integrated models to avoid tipping points in a multi-objective water allocation problem

Abstract:

Nonstationary environmental change is exacerbating existing water scarcity, stressing interdependent human and environmental systems. Sustainably managing water for both ecological and agricultural uses under these changes is a critical challenge for climate adaptation. This challenge is further complicated by the presence of interacting physical, biological, and socioeconomic tipping points governed by uncertain dynamics. Managing these systems for multiple objectives, then, seems to require integrated models that jointly consider policy decisions and their effects on ecological and agricultural systems. Embedding such an integrated model within an adaptive management problem would further allow a manager to accommodate and reduce uncertainty around a solution to this challenging multi-objective problem. Here, we leverage recent progress in robust control to solve an adaptive management problem using an integrated model with water-dependent agricultural and ecological systems. In this model the ecological system exhibits threshold-like behavior and the water allocation policy reacts discontinuously to the state of the ecological system, which reflects the potential for changes in threatened species status. We explore how managing this system while learning its behavior becomes more complex in scenarios with greater uncertainty. Our focal scenario includes unknown stationary parameters defining the tipping point, forced by an unknown non-stationary stochastic process. Using an ecological model for salmonid species, we examine the effects of single- versus multi-objective management in this system. We demonstrate: (1) Adaptively managing for a single objective without consideration of the integrated system can lead to adverse outcomes; and (2) that robust control provides a flexible framework for exploring patterns of effective management strategies in integrated systems with different levels of uncertainty.