Science1 publisher2 min readPublished
Hidden funnels let simulated fast-slow systems settle where simpler models say they cannot
Serhiy Yanchuk's team at University College Cork found narrow 'funnels' extending from stable-state basins in simulations of up to 10 linked oscillators. Because the work rests on model systems, how far the funnels alter real climate or brain forecasts is an open question.
The Scientist · Science desk
What happened
- The funnels are long, narrow pathways stretching out of some basins that let a system return to a stable state from starting points far beyond where it would be expected to.
- The team built models that pair fast and slow processes, mapped which starting conditions led to each stable state, and simulated how the systems evolved.
- Funnels narrow dramatically as the gap between fast and slow timescales widens, but they never fully disappear while both processes run at finite speeds.
- Funnels appeared in several systems, starting with the simplest possible model of two competing states, and that spread suggested they could be a universal feature.
- The work, by researchers in Ireland and Germany, appears in Physical Review Letters.
Compiled by The ScientistSomething wrong?How this is made
Why it matters
- constraint Anyone judging resilience or tipping points from a simplified model alone, including in climate-impact work, has to allow that fuller fast-slow dynamics could send a disturbed system to a different state.
- precedent If the funnels prove as general as their spread across test systems suggested, basin maps of other multiscale models would need to be checked for them.
- capability Modellers now have a specific structure to look for, and according to the account, models that include it should predict change in multistable systems more reliably.
A basin of attraction is the full set of starting points that eventually lead to a given stable state [3]. Resilience questions start from it: knock a system away from its state, then ask whether it returns or tips into a different one [15]. The paper's title calls the funnel effect "tunneling between stable states" [13]. Put plainly, a system can reach a stable state from a starting point that a simpler model would rule out entirely [2].
This is a tidy result, and the part I find most useful is how the funnels change with the timescale gap. The account's example of mixed timescales is the climate, where fast-changing weather is coupled to far slower changes in oceans and ice sheets [9]. If the team's trend holds there, a gap that wide is where the funnels would be thinnest [1].
The thing this doesn't tell you is how often a real disturbance would land in a funnel. The phys.org account does not report funnel widths or the fraction of starting conditions they occupy. Any estimate of tipping risk would need that fraction as its denominator.
The account's central claim is that a disturbance could switch a system between states in a way simplified models cannot capture [7]. Brain signals and climate patterns appear in the account as examples of systems with more than one stable state [14]. The comparison itself was made inside the team's own fast-slow models [4]. To check whether a funnel changes the tipping behaviour of a working climate or neural model, someone would have to build a basin map of that model.
What to watch
- Published funnel widths, or the share of starting conditions the funnels occupy, for systems with realistic timescale gaps.
- A basin map of an established climate or neural model that either finds singular funnels or fails to.
- Tests of whether the funnels persist in networks much larger than 10 oscillators, or when the system is under random forcing.