resilience-memory-utac

GenesisAeon Package 97 — real critical-slowing-down statistics and post-perturbation resilience-debt tracking. Deliberately has no UTAC/CREP/AFET bridge — see DISCLAIMER.md.

For a plain-language explanation of the same topic (German, no jargon, written for general audiences), see WHITEPAPER.md.

What's real here

• Dakos et al. (2012, PLOS ONE): the generic early-warning-signal workflow — detrend a time series, compute rolling variance / lag-1 autocorrelation / skewness, test each for a trend with Kendall's tau.
• Scheffer et al. (2009, Nature): the underlying concept — recovery rate from small perturbations declines as a system nears a critical transition ("critical slowing down").
• van Nes & Scheffer (2007, American Naturalist): slow recovery from perturbations, tracked here as an explicit "resilience-debt" — if a system hasn't recovered before the next shock, it carries the deficit forward.
• A live, verified illustration (not calibration input): Rhine at Kaub (24 cm, 2026-08-04, matching the 2018 record), Danube at Paks (-144 cm, below the 2018 record of -98 cm, triggering Hungary's first-ever full reactor shutdown), and Cernavoda NPP (Romania, reactor shutdown + emergency underwater demolition to redirect cooling water) — see DISCLAIMER.md for exactly what this case study is and is not used for.

Quickstart

pip install resilience-memory-utac

from resilience_memory_utac import compute_early_warning_report

# any real time series you supply -- river discharge, coral cover, ...
report = compute_early_warning_report(my_series)
print(f"Variance trend: tau={report.variance_tau:.2f} (p={report.variance_p:.3f})")
print(f"Autocorrelation trend: tau={report.autocorrelation_tau:.2f}")

from resilience_memory_utac import PerturbationEvent, compute_resilience_debt

events = [
    PerturbationEvent(label="2018 low-water", time=0, magnitude=8.0, tau=365.0),
    PerturbationEvent(label="2026 low-water", time=2800, magnitude=10.0, tau=365.0),
]
states = compute_resilience_debt(events)
print(states[-1].fully_recovered)  # False if the 2018 event hadn't fully decayed away

Development

pip install -e ".[dev]"
pre-commit install
ruff check src tests
mypy src
pytest

Citation

See CITATION.cff and .zenodo.json.