Detecting extreme event-driven causality

The occurrence of certain extreme events (such as marine heatwaves or exceptional circulations) can exert causal influences on subsequent extreme events (such as heatwave, drought and flood). These concurrent extreme events have a profound impact on environment and human health. However, how to detect and quantify the causes and impacts of these extreme events through a data-driven way remain unsolved. In this study, Dynamical Systems approach is extended to develop a method for detecting the causality between extreme events. Taking the coupled Lorenz-Lorenz systems with extreme event-driven coupling as an example, it is demonstrated that this proposed detection method effectively captures extreme event-driven causality, exhibiting improved performance in detecting causality between concurrent extreme events. This study also examines the impact of complete versus partial observations on causal inference performance and demonstrates that the embedding technique can improve the accuracy of causal detection. The successful application to the Walker circulation phenomenon demonstrates the generalizability of our method and provides a novel contribution to causal inference research in complex systems. This method offers valuable insights into multi-scale nonlinear dynamics, particularly in revealing associations among extreme events.