From raw climate data to a single number per cell.
Every reading on the Climate Mitigation Atlas — every β value, every "no data" cell, every best-pathway call — comes out of the same five-step pipeline. This section walks through the pipeline with a concrete worked example: the permafrost finding at +0.36 K.
The pipeline in one sentence
Take a climate observable trajectory \(\Phi(t)\) → finite-difference its rate \(\rho(t) = \mathrm{d}\Phi/\mathrm{d}t\) → fit \(\beta\) from \(\log|\rho|\) against \(\log|\Phi|\) → compare \(\beta\) across structurally different model trajectories → publish the cross-model dispersion \(\sigma_{\text{cross}}\) alongside a four-tier verdict and an explicit scope-reporter.
Five steps, all data-driven
Step 1
Ingest the data →
CMIP6 model output via Pangeo cloud catalogue (NetCDF/zarr). Observational records (NSIDC, RAPID, CERES, NOAA NCEI). Paleoclimate proxies (Vostok, EPICA, NGRIP). No domain-specific pre-processing.
Step 2
Apply the operators →
\(\mathcal{B}\) extracts β per model. \(\mathcal{S}\) measures cross-model dispersion. \(\mathcal{M}\) computes consensus. \(\mathcal{P}\) decomposes cadence-stamped cascades. PELT detects regime-change points. No imported thresholds.
Step 3
Project forward →
Statistical fans (ARIMA + rate perturbations) for observational records. CMIP6 SSP fans for forced projections. Three brake estimators in parallel: polyfit, GLS-AR(1), Bayesian-AR(1). Cross-checked.
Step 4
Synthesise the verdict
Four-tier vocabulary: robust, partially-robust, inconclusive, theorem-3-bound. Plus Outcome A/B/C for paleoclimate regime changes and "model-artefact" for outlier flagging.
Step 5
See it end-to-end on permafrost →
A complete worked example: from raw CMIP6 cSoil files for 6 ESMs across 5 SSPs, through brake-extraction and PELT change-point detection, to the cross-model warming-anomaly of +0.36 ± 0.12 K that the framework reports as "already crossed". Every command. Every dataset. Every script.
What the framework does NOT do
The pipeline is intentionally opinionated about scope. Under Law V (Honest Scope):
- It does not import any climate-domain threshold (no °C cutoff, no ppm target). The only universal threshold is β = 1, drawn from the framework's first theorem.
- It does not tune per observable. Same operators, same code, same statistical tests are applied to atmospheric CO₂, sea-ice, AMOC, Amazon NPP, GMSL, permafrost, energy budget, and ozone.
- It does not hide where it doesn't apply. Where Theorem 3 binds (narrow log-range), the framework reports inconclusive. Where Theorem 8 fires (forced-rate degeneracy), the framework reports β ≈ 0 by construction. Where Theorem 10 fires (anti-shadow detection), the framework declares the candidate set jointly inadmissible.
Reproducibility
Every experiment writes a TXT human summary + JSONL payload to domains/climate/results/. All scripts under domains/climate/experiments/. The framework apparatus (operators, theorems, dual-number representation) is at uniformity.senuamedia.com/framework/.