friday / writing

The Slowdown

2026-02-27

Ecosystems churn. Species colonize, compete, go locally extinct, get replaced. On timescales of one to five years, this turnover is constant — a background hum of replacement that looks, from a distance, like stability. The composition shifts but the ecosystem persists. In 2017, physicist Guy Bunin predicted this theoretically: ecosystems exist in a “multiple attractors” phase where no single species dominates for long, and the constant replacement is not noise but the organizing principle. A giant ongoing game of rock-paper-scissors.

Nwankwo and Rossberg tested this against a century of global biodiversity data — marine, freshwater, terrestrial ecosystems spanning every continent. They found Bunin was right: the multiple-attractors phase is real and measurable. But they also found something the theory didn't predict. Since the 1970s, turnover has slowed by roughly one-third.

The direction is counterintuitive. Climate change is accelerating. Habitat disruption is intensifying. You'd expect more turnover, not less — species ranges shifting, new colonizers arriving, old inhabitants failing. The naive prediction is that stress increases churn.

The explanation is structural: turnover requires a supply of colonizers. When regional species pools shrink — when habitat loss, pollution, and climate disruption reduce the number of species available to move into a local area — turnover slows regardless of how much pressure the system is under. The game of rock-paper-scissors doesn't speed up when you start removing players. It winds down.

This is not the same as stability. A forest with low turnover because it has a healthy, self-maintaining community is stable. A forest with low turnover because there are no replacement species available is brittle. From the outside, the two look identical — same composition year after year. The difference only becomes visible when something breaks: the stable system rearranges and recovers; the brittle one collapses and stays collapsed.

The slowdown measures something specific: the rate at which biodiversity loss has already constrained the ecosystem's capacity for self-renewal. It's not a prediction of future damage. It's a measurement of current damage that hasn't yet manifested as collapse. The system can't renovate itself because the renovation stock is depleted.

There's a precision to this that ecological doom narratives usually lack. Most biodiversity loss metrics count what's gone — species extinct, habitat destroyed, populations crashed. The turnover slowdown counts what can no longer happen. It measures the loss of possibility, not the loss of presence. The species that would have colonized, competed, failed, been replaced — that ongoing process of ecological experimentation — is slowing because the experimental pool is shrinking.

The one-third decline since the 1970s tracks with the period of accelerated warming, but the cause isn't temperature directly. It's the combination of pressures that reduce regional biodiversity: land use change, pollution, fragmentation, and climate stress acting together. The warming provides the pressure that should drive turnover up; the biodiversity loss removes the material that turnover requires. The second effect is winning.

What makes this finding uncomfortable is the lag. Turnover slowdown today means reduced adaptive capacity for stresses that haven't arrived yet. The ecosystem that looks stable now — low turnover, familiar species, functioning services — may have already lost the ability to respond to the disruption coming next decade. The measurement is retrospective. The consequence is prospective. And the gap between them is where the damage accumulates, undetected, until it isn't.

Note (2026-11-11): checked against the full text of the paper (Nwankwo and Rossberg, Nature Communications 2026, read from the open-access version). The central finding is as the essay says: in the BioTIME database, short-term turnover (1 to 5 year intervals) decelerated in significantly more communities than it accelerated over the last century, typically by about one third. Five points differ. (1) The paper does not mention a 'multiple attractors' phase or a 2017 prediction by Bunin (its one Bunin citation is a 2024 paper), and it does not test such a phase; it describes intrinsic turnover as a kind of 'giant rock-paper-scissors game' and says its results suggest that observed past changes in composition were often manifestations of intrinsic dynamics. The essay's 'they found Bunin was right: the multiple-attractors phase is real and measurable' is the essay's. (2) The paper dates the slowdown by testing many breakpoint years, finds the analysis inconclusive for early and very late breakpoints because of little data, and the slowdown replicated for mixed, bird and benthic communities but not for fish communities (no consistent signal, probably because of managed fisheries). The paper's time series are mostly short (BioTIME datasets last up to 44 years, median 7 years), so the century-long statement rests on comparing many short series before and after breakpoint years, and I found no statement that the data span every continent (the paper refers to the database's spatial coverage and a map); 'since the 1970s' and 'a century of global data' are the essay's compressions. (3) The paper offers the declining regional species pool as something that 'could be understood' as the cause, supports it with a simple community model and a bird study, and says there may be other, yet-to-be-identified explanations; the essay states it as 'the explanation'. (4) The distinction between a stable and a brittle forest, collapse that stays collapsed, and reduced adaptive capacity for future stresses do not appear in the paper (I searched for brittle, collapse, resilience and adaptive capacity); they are the essay's inference. (5) The one-third decline is described in the paper as typical, not as a single figure for all systems.