Status
Partial — ~24% complete
6/25 model-SSP pairs complete due to the heavy ocean-heat-content integration. Partial results show closure ratios of 0.397 to 0.549 across SSP1-2.6 to SSP5-8.5 — broadly consistent with instance #18's observational 78% closure. The full fan is queued behind streaming gcsfs OHC fetches.
Pattern-4 framework
Earth-system energy conservation: \(\sum_i \Phi_i = \Phi_{\text{total}} + \text{boundary}\), where \(\Phi_i\) are reservoir energy stocks (ocean, ice, land, atmosphere) and \(\Phi_{\text{total}}\) is integrated TOA net flux. Each reservoir is an independent component cascade (NOT a shadow of the others). Partition fraction \(f_i(t) = \rho_i(t) / \sum_j \rho_j(t)\).
Observational anchor (instance #18)
Time-mean closure ≈ 78%. The 22% gap matches deep-ocean (>2000m) + deep-cryosphere reservoirs the four measured shadows do not capture. Within-atmosphere σ_cross over three temperature datasets = 0.025 (within-shadow noise = 0.07).
PELT detects coherent 2012–2014 regime shift in 4/5 reservoirs (ocean 2013, ice 2012, TOA 2012, atmosphere 2014; land alone has no CP). Framework reads this as a regime shift event, not steady drift.
Source: scenario_fan_energy_budget_v4_chunked.py + conservation_decomposition_v2.py