Per-SSP permafrost cSoil first-PELT change-point

SSPn_models OKfirst_cp medianfirst_cp IQRwarming@cp median (K)sink→source flips
SSP1-1.912028[2028, 2028]+0.2710/1
SSP1-2.662033.5[2029, 2034]+0.5172/6
SSP2-4.562029.5[2028, 2033]+0.4751/6
SSP3-7.062028.5[2028, 2029]+0.3182/6
SSP5-8.562027.5[2027, 2029]+0.3630/6

First-break warming threshold crossed in 2026

Cross-model warming-anomaly at first PELT change-point: median = +0.36 ± 0.12 K across all 6 models under SSP5-8.5. Far below the published +1.5–2.0 °C tipping window. Already crossed at 2026 global mean temperature anomaly (~+1.2 K above pre-industrial). The framework reads first regime change as having already happened.

Best mitigation pathway — SSP1-2.6

SSP1-2.6 delays first PELT-CP to median 2033.5 (vs SSP5-8.5 at 2027.5). ~5 yr gain from aggressive mitigation. SSP1-1.9 has only 1 model in the cache (CESM-WACCM not yet run) so the median is unreliable.

Sink → source flip non-monotonicity

Across the 5 SSPs:

  • SSP1-1.9: 0/1 models flip sink → source (n=1)
  • SSP1-2.6: 2/6 models flip
  • SSP2-4.5: 1/6 models flip
  • SSP3-7.0: 2/6 models flip
  • SSP5-8.5: 0/6 models flip

Sink→source flips are not monotone in forcing. Highest mitigation (0/1) and highest emissions (0/6) both yield zero flips. Intermediate SSPs show 1–2 flips. Framework interpretation: under SSP5-8.5, high warming + high productivity simultaneously boost soil-C influx; net sign holds. Under aggressive mitigation, the cascade stays small and flip detection becomes noise-limited.

cSoil brake-p β values themselves bind against Theorem 3 (median log-range only 0.012–0.018) — Layer-A inconclusive. The first-break warming threshold is the operative finding, not β.

Source: scenario_fan_permafrost.py