โ† Decision Making and Prediction

Early-warning indicators

Early-warning indicators are measurable features intended to signal increasing risk before a major transition or undesirable outcome occurs.

Possible indicators

Depending on the system, indicators may include changing growth rates, increasing variance, increasing autocorrelation, altered spatial patterns or probabilities derived from a fitted stochastic model.

Critical slowing down

Near some dynamical transitions, recovery from perturbations becomes slower. This can produce increased temporal persistence and variance, although these signals are not universal.

Validation

An indicator is useful only if it gives sufficiently reliable and timely information for the intended decision. False alarms, missed events and lead time should therefore be evaluated.

Key idea. An early-warning signal is decision-relevant only when its ability to anticipate the target event has been tested.