Assessing Energy-Water Dynamics with Scalable Agent-Based Modeling Approaches

With the growing concern about the connection between energy and water systems, process-based modeling is a popular method to evaluate energy-water (EW) dynamics and assess potential non-stationarity under future conditions. However, a common limitation for these types of models is the adoption of a fixed scale across complex system components and dynamics, which could lead to misrepresentation of metrics characterizing resilience. Usually, an energy-water coupled-natural-human system will be identified first (e.g. a city, a basin, a state or a country) and a specific water-energy trade off model will be developed for such system. Under the assumption that interactions are similar, these case study models might be generalizable such that they can be applied to other similar systems with available data. However, such models cannot be applied to other scales due to lack of a systematic approach to address the scalability of human decision uncertainty inside the modeling structure. This presentation summarizes the progress of the first two years of a DOE-funded project that focus on developing and utilizing a scalable agent-based modeling approach to address this research gap. Using the Colorado River and its subbasin, the San Juan River, as examples, we will demonstrate how to use a Bayesian Inference mapping approach combined with the Cost-Loss method to quantify risk aversion among geographically defined agents at different spatial scales. We define agents as groups of water users, water regulators, energy users and energy producers. Capturing an agent's risk aversion plays a key role in overall EW system's operation and affect EW system-wide performance and resilience—as well as the resilience of interconnected systems such as the electric grid. The scalable agent-based model also allows us to explicitly test and evaluate the adoption of adaptation measures. This includes the adoption of new technology or changes in the way that agents operate, both of which are complicated by the perceived position of the agent with respect to future climate and how water/energy/food markets are evolving (e.g., differences in utility, uncertainty, and risk aversion parameters) across scales.