amoc-utac
GenesisAeon Package 18 — Atlantic Meridional Overturning Circulation as UTAC System
AMOC modelled as UTAC dynamical system with physics-based early-warning from van Westen et al. (2024).
Key result: Γ_AMOC ≈ 0.251 (medium-CREP). Neural criticality (Package 20) lands at the same Γ because it shares the same η = 50 % and σ = 2.2 calibration inputs — a shared comparison coordinate, not evidence of shared physics between the two domains.
Installation
pip install amoc-utac
For development:
pip install -e ".[dev]"
Quickstart
amoc-utac run --duration 120
amoc-utac tipping-estimate
amoc-utac zenodo-export
Integration in genesis-os
from genesis_os import GenesisOS
os = GenesisOS()
amoc = os.load_package(18)
results = amoc.run_cycle(duration_years=120)
Benchmark
Validated against RAPID array, van Westen 2024 & Ditlevsen 2023.
Falsifiable Prediction
AMOC crosses 50 % weakening (H* = 0.5 K) between 2045–2065.
Scientific Context — Consensus vs. Single Studies (added 2026-08-01, corrected 2026-08-02)
The Ditlevsen & Ditlevsen (2023) tipping-year estimate this package compares
against is a real, published, but more alarmist single study, not the
mainstream consensus. Important: the paper was itself corrected in 2025
(Author Correction, Nat. Commun. 16, 7794, DOI 10.1038/s41467-025-63201-y)
— the central estimate moved from 2057 [2025-2095] to 2065 [2037-2109];
this package now uses the corrected values. A 2026 preprint (Morr et al.,
arXiv:2604.20341, not yet peer-reviewed) further argues the underlying
statistical fingerprint/model choice is fragile enough that alternative,
equally-defensible specifications push the estimate much later.
The IPCC AR6 Working Group I Summary for Policymakers (2021, Section
C.3.4) states: "there is medium confidence that the Atlantic Meridional
Overturning Circulation will not collapse abruptly before 2100." Note this
is not the same as the commonly-repeated shorthand "very unlikely before
2100" — "very likely" in the same AR6 passage describes the assessed
weakening itself (24% [4-46%] under SSP1-2.6, 39% [17-55%] under
SSP5-8.5 by 2100), not an exclusion of collapse, and "medium confidence" is
not a numeric probability. Both the corrected Ditlevsen estimate and the
AR6 consensus position are surfaced together in predict_tipping_year()'s
return value (ipcc_ar6_confidence_statement, ipcc_ar6_citation,
consensus_note) and in the tipping-estimate CLI command's output table,
so users see the consensus baseline alongside this package's own (more
alarmist) UTAC estimate rather than only the latter.
Global Trajectory Context (added 2026-09-02)
UNEP's "Limiting Overshoot: Navigating Exceedance of 1.5°C and Pathways Towards Return" (UNEP, published 2026-09-02, DOI: 10.59117/20.500.11822/49857) assesses that current policies put the world on track for ~2.6°C warming by 2100, with permanent exceedance of the 1.5°C Paris threshold expected within the next few years (2024 was already ~1.6°C above pre-industrial levels, per Copernicus/C3S). The report's central pathway is "overshoot, peak, and decline": exceed 1.5°C, then attempt to bring temperatures back down via large-scale carbon dioxide removal.
What this does and does not change here: this package's own results
(Γ_AMOC ≈ 0.251, the corrected Ditlevsen tipping-year estimate, the AR6
SSP1-2.6/SSP5-8.5 weakening ranges) are not altered by this report — UNEP
does not publish an AMOC-specific estimate. What it does provide is
current, authoritative context for which emissions scenario is
realistic: a ~2.6°C-by-2100 trajectory sits closer to the SSP5-8.5
end of the AR6_WEAKENING_SSP585_PCT range in constants.py than to
SSP1-2.6, which is worth keeping in mind when interpreting this
package's output against the AR6 consensus band.
Role in the GenesisAeon Ecosystem
amoc-utac is P18 in the GenesisAeon ecosystem, covering the
oceanography / AMOC tipping domain. It applies the UTAC-Logistic ODE and
CREP metrics to the Atlantic Meridional Overturning Circulation, sharing
its criticality framework (Γ ≈ 0.251) with sibling packages across other
domains (e.g. neuroscience, heliophysics).
Citation
License
Code: MIT • Docs & Data: CC BY 4.0
Metadata
Release files for amoc-utac 1.3.3
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| amoc_utac-1.3.3.tar.gz | 210.7 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| amoc_utac-1.3.3-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 240.8 kB
Release files / amoc_utac-1.3.3.tar.gz
| Download URL | amoc_utac-1.3.3.tar.gz |
|---|---|
| Size | 210.7 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
5f4de6e831300cb6119fc1f1899a30c766a65095b5239df7e179a751ab9c89e4
|
|
BLAKE2b-256 checksum How to use checksums |
9656f9e2c8f182609c3d68842b382218da59b0c60b401e069e1bb839528c53d1
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Release files / amoc_utac-1.3.3-py3-none-any.whl
| Download URL | amoc_utac-1.3.3-py3-none-any.whl |
|---|---|
| Size | 30.1 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
374eaa1467b783fc7b11ba0e9fe939d9b26dd109e3fe85c045c37fbaea6d160b
|
|
BLAKE2b-256 checksum How to use checksums |
5b761fece04aa7647b10a297c7c3e80f8d13174ccdb2ca67b5a481b2045f1941
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/7.0.0 CPython/3.13.14
|
Security for amoc-utac 1.3.3
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.