Three things we can change. Three we can't.

Climate-tipping science has been hampered by a single technical problem: model ensembles produce wide tipping-threshold ranges (AMOC: 1.4–8°C; Amazon: 2–6°C) because the models disagree. The framework's contribution is a unified diagnostic that tells you, observable by observable, which signals are robust across models — and therefore which mitigation choices actually matter.

What we can change

Stoppable

CO₂ growth itself

Under SSP1-2.6 the framework reads atmospheric CO₂ as deeply sub-rate post-2024 (β = −24.3). Under SSP5-8.5 it crosses β = 1 forward in 2038. Mitigation works on the trajectory.

Stoppable

Aggregate sea-level rise

SSP1-1.9 wins the framework's three-criterion best-pathway test (rank-sum 7 of 24): tightest cross-model agreement, smallest mean rate, β farthest below 1.

Stoppable

Glacier decline rate

Glaciers stay sub-rate everywhere, but the rate scales with forcing (β = 0.21 SSP1-1.9 → 0.90 SSP5-8.5). Mitigation buys you decades of glacial retention.

What we can't

Locked in

Antarctic ice-sheet collapse

β = 3.21 to 3.76 across every SSP including SSP1-1.9. AIS dominates sea-level rise post-tipping; aggregate GMSL projections hide the component lock-in.

Locked-in risks →

Already crossed

Permafrost first regime change

The framework identified +0.36 ± 0.12 K as the first permafrost change-point. Earth is at +1.2 K. The threshold is in the past tense. Mitigation now reduces post-tipping rate, not whether the tip has happened.

Locked in

Greenland ice-sheet acceleration

β = 1.18 to 1.51 across all SSPs. Less dramatic than Antarctica, but the same conclusion: super-rate under every emissions pathway. Mitigation reduces magnitude, not regime.

What's uncertain — and why the answer surprised us

Counter-intuitive

Aggressive mitigation widens, not narrows, the uncertainty band on sea-ice and AMOC

Under SSP1-1.9, model disagreement on Arctic sea-ice (σ_cross = 5.10) is 7× wider than under SSP5-8.5 (σ_cross = 0.69). The forcing signal disappears beneath internal variability. The same effect appears for the AMOC.

Standard ensemble-mean climate projections do not surface this — they take per-SSP means and don't measure cross-model dispersion per SSP. Aggressive policy assessments need wider uncertainty bands, not narrower.

Read the full paradox →

Two more findings worth your attention

Aggregate hides truth — focus on components

Sea-level rise projections look reassuring under SSP1-1.9 (aggregate β = -1.12, sub-rate). The dominant component — Antarctica — is super-rate (β = +3.76) under the same SSP. The framework's component decomposition reveals where mitigation actually matters: the Antarctic-specific feedback loops (basal melt, marine ice cliff instability) drive the post-tipping rate, not aggregate emissions.

Component decomposition →

Montreal worked. The original intervention was the lever.

Counterfactual analysis: without the 1987 Montreal Protocol, the framework reads Antarctic ozone hitting zero by 2010. Strengthening compliance further produces only marginal recovery acceleration. The original act of intervening was the principal lever, not subsequent tightening. A useful template for what successful intervention looks like in framework-native terms.

Ozone counterfactual →

Read deeper