Effects of noise correlation and imperfect data sampling on indicators of critical slowing down

Abstract

Critical slowing down-based early warning signals (EWSs) are well-known indicators that precede an approaching collapse in complex systems. To date, the majority of studies on the predictability of critical transitions consider systems perturbed with temporally uncorrelated noise. In contrast, here we study catastrophic and non-catastrophic transitions, and the performance of associated EWSs in systems perturbed with correlated noise. We find that elevated noise correlation can advance the occurrence of a catastrophic transition, and simultaneously progresses the system’s recovery. However, noise correlation does not have a significant impact on the likelihood of non-catastrophic transitions. We show that depending upon the transition mechanism, the occurrence of weak to false signals increases with noise reddening. Imperfect data sampling, both spatial and temporal, further reduces the efficacy of EWSs. Spatially limited data has more impact on the efficacy of EWSs for negative noise correlation than that of positive. However, temporally imperfect data is more detrimental for positively correlated noise. Overall, our study suggests that performance of EWSs is critical to system-specific perturbations as well as data sampling.