Regensburg 2025 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

DY: Fachverband Dynamik und Statistische Physik

DY 27: Poster: Nonlinear Dynamics, Pattern Formation, Granular Matter

DY 27.3: Poster

Mittwoch, 19. März 2025, 15:00–18:00, P4

Causal inference in nonlinear Covid-19 time series — •Adrian Pelcaru and Dirk Brockmann — SynoSys/TU Dresden, Dresden, Germany

Covid-19 pandemic exemplifies a spatio-temporal dynamic system which can hardly be studied by controlled experiments with "laboratory settings". Advances of computational power and availability of large datasets, causal discovery frameworks enable one to gain insight into the dynamics and couplings between observables of the system. We employ an approach rooted in state space reconstruction called Convergent Cross-Mapping (CCM) and investigate time-lagged interactions between multiple observables (temperature, interventions, human mobility, reproduction rate) of Covid-19 in 114 countries and 38 regions in Germany. Central, is the interrelationship between human mobility and reproduction rate. We find evidence for time lagged feedback couplings between human mobility and reproduction rate and identify other drivers of the reproduction rate acting either directly or indirectly through human mobility. Furthermore, we detect latitudinal dependence of coupling strengths and clear segregation of coupling strengths between historic east and west Germany. Finally, we measure dynamic tipping points of strong unidirectional forcing between human mobility and reproduction rate leading to distinct periods with characteristic dynamics. While many causal discovery frameworks and previous studies related to Covid-19 a priori reject potential feedback mechanisms between observables of a system, this study, to our knowledge, shows their existence for the first time.

Keywords: COVID-19; Causal Inference; Time Series; Empirical Dynamic Modeling