Regensburg 2022 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
SOE: Fachverband Physik sozio-ökonomischer Systeme
SOE 6: Data Analytics for Complex Systems (joint session DY/SOE)
SOE 6.3: Talk
Monday, September 5, 2022, 15:30–15:45, H18
Bayesian approach to anticipate critical transitions in complex systems — •Martin Heßler1,2 and Oliver Kamps2 — 1Westfälische Wilhelms-Universität Münster, 48149 Münster — 2Center for Nonlinear Science, Westfälische Wilhelms-Universität Münster, 48149 Münster
Complex systems in nature, technology and society can undergo sudden transitions between system states with very different behaviour. In order to avoid undesired consequences of these tipping events, statistical measures as variance, autocorrelation, skewness and kurtosis have been proposed as leading indicators based on time series analysis. Under favourable conditions they can give a hint of an ongoing bifurcation-induced destabilization process. However, they suffer from their loose connection to complex system dynamics, sensitivity to noise and sometimes misleading trends. Therefore, we want to present an alternative approach assuming the dynamical system being described by a Langevin equation. Starting from this stochastic description, we combine MCMC sampling, rolling window methods and Bayesian reasoning to derive the drift slope as an alternative early warning sign. The Bayesian approach enables us to define credibility bands which make it easier to distinguish random fluctuations from real trends that imply a less resilient system. Our investigations suggest that the estimation procedure is rather robust even under strong noise. Besides, the noise level of the system is computed to get insights into the probability of a noise induced transition. We want to present some of the results and discuss possible limitations and tasks of future research.