Early Warning Signals and Noise Design

Systems can undergo gradual changes that significantly impact their stability. Due to these gradual changes, a system can show a sharp switch between contrasting states. This sharp switch is also known as a tipping point event. For example, ecosystems can be exposed to gradual changes of the environment such as changes in pollutants and available nutrients. In turn, these gradual changes might result in abrupt bursts or extinctions of species in the ecosystem. Identifying when a system is close to a tipping point is a fundamental question to be able to prevent potential catastrophic effects. One can obtain signals from observed data of components of a system to identify that a tipping point event is about to occur. These signals are called early warning signals (EWSs) and are the main topic of this talk. 

We want to investigate EWSs in the context of networked systems, where we can account for the interactions between the components of the system. For example, in an ecological network we can describe how some species predate on others. These systems can be exposed to external noise such as the noise exerted by the environment on ecological networks. Our research question is how to further principled methods to exploit external noise perturbations and produce "good" EWSs. In this talk I will show preliminary work on this research question.