Probabilistic Prediction of the AMOC Tipping Point (UCD-AMOC)

Prédiction Probabiliste du Point de Bascule de l'AMOC (UCD-AMOC).pdf

Télécharger PDF • 6.74MB

Probabilistic Prediction of the AMOC Tipping Point (UCD-AMOC).pdf

Télécharger PDF • 3.56MB

The Universal Dynamic Convergence Formula for Probabilistic Prediction of the AMOC Tipping Point (UCD-AMOC) is a probabilistic framework adapted from the Stommel model, designed to model the collapse of the Atlantic Meridional Overturning Circulation (AMOC) under noise and bistability.

It integrates parameters such as salinity thresholds (S), hydraulic resistances (R), ocean volumes (V), flow divergences/convergences (D/C), time cycles (T), dynamic memory (M), energy density (E), observables O(t) and drivers D(t), via a function φ normalized by A.

Simulations with odeint and Monte Carlo show probabilities C(t) from 0 to 1, with tipping at gamma >1.5, RMSE <0.08, and Lyapunov stability, validated on RAPID-compatible simulated data (2025). The Python implementation (NumPy, SciPy) includes calibration via XGBoost, with graphs (Figures 1-13) illustrating sensitivity, robustness and comparisons to models such as WRF, ECMWF or CMIP6. Limitations include the need for empirical validation (NOAA/ECMWF) ; perspectives : extension to bioclimate, AI for real forecasts, and submission to arXiv for peer-review.