Source
Scheffer, Bascompte, Brock, Brovkin, Carpenter, Dakos, Held, van Nes, Rietkerk and Sugihara (2009), LIT-1956, read in NOTE-1536 from chapter 2 (pp. 17–32) of V. Dakos, Expecting the Unexpected (Wageningen, 2011), edepot.wur.nl/168815. Earlier work: Scheffer et al. 2001 (LIT-720) and Scheffer and Carpenter 2003 (LIT-726).
What was actually shown
Slowing down (p. 19). “At fold bifurcation points … the dominant eigenvalue characterizing the rates of change around the equilibrium becomes zero”, which “can be proven” for “any continuous model approaching a fold bifurcation” (Wissel 1984). Box 2.2’s transcritical example gives λ₁ = −γ(b − a), zero at b = a (checked).
Autocorrelation and variance (pp. 21–22). For the linearized stable point dx = λx dt + σ dW, λ < 0, the stationary variance is σ²/(2|λ|) and the lag-Δ autocorrelation e^{λΔ}; both rise without bound or to 1 as λ → 0⁻. Simulated here at λ = −2, −1, −0.5, −0.1: variances 0.26, 0.53, 1.03, 5.16 (exact 0.25, 0.50, 1.00, 5.00), autocorrelations 0.14, 0.40, 0.62, 0.91 (exact 0.14, 0.37, 0.61, 0.90). The review grants that variance could in principle fall as the system tracks fluctuations less well, but models show it “usually” rises (p. 21).
Signals without slowing (pp. 22–24). Skewness, from the unstable point approaching on one side; flickering, noise-driven switching once bistable.
The scope (p. 31). “The same signals may even occur, albeit in a less pronounced way, as the system approaches a threshold that is not related to catastrophic bifurcations”, so they indicate “proximity to a broad class of thresholds, where small forces can cause major changes”. And “perturbations will often trigger a transition well before a bifurcation point is reached”, so the moment “remains difficult to predict”.
What this does not say
- That the signals are reliable in real data: detection needs long series and filtering choices, and false positives and negatives are listed (p. 31).
- That they are known for chaotic systems or basin-boundary collisions: “we know of no explicit work” on the latter (p. 24).
- That slowing near a transition makes new organization. It marks proximity to a threshold, of either kind.