Predicting Tipping Points in Water Supply using Early Warning Signals

Abstract:

Population growth and a drying climate can push urban water supply across a tipping point into a new regime of water deficits. Critical transitions in dynamic systems may be preceded by critical slowing down, where the system recovers from small perturbations at a slowing rate. Statistical properties of system dynamics may show significant changes during critical slowing down and can serve as Early Warning Signals to detect the approach and occurrence of a tipping point. In this research, we develop a comprehensive modeling framework to detect critical transitions in water supply and test management strategies for averting tipping points using Early Warning Signals. We use an agent-based modeling framework of water use, water supply, and water management to simulate feedbacks and adaptations between human behavior and water infrastructure systems. Projections of climate change and population growth are simulated to apply increasing levels of stress on the water supply system. Climate change scenarios are generated using a stochastic reconstruction framework, which models shifting streamflow, precipitation, and evapotranspiration to create stress gradients. This work applies the modeling framework for an illustrative water supply system over a range of stress gradients for a projected period of 100 years. Seven Early Warning Signals are computed for log-transformed reservoir storage, and these indicators increase in anticipation of a tipping point, indicating critical slowing down for some climate change scenarios. Thresholds in Early Warning Signals indicators are developed to match the observed occurrence and timing of tipping points, which are confirmed through inspection of hysteresis curves. Management strategies are evaluated based on the reduction in the expected number of tipping points and the delay in critical transitions. Results demonstrate the use of the framework for predicting and averting tipping points in water supply management.