Source Code and Demo for multi-decadal AMOC prediction model

### OverviewThis repository provides a Convolutional Neural Network (CNN) based AI emulator designed for projecting Atlantic Meridional Overturning Circulation (AMOC) strength anomalies. The model is developed to facilitate rapid and accurate assessments of AMOC stability by integrating multi-modal climate input fields. ### Key Features* **AI Architecture:** A multi-input CNN framework optimized for spatio-temporal climate data.* **Input Variables:** Incorporates Surface Heat Flux (SHF), Surface Freshwater Flux (SFF), Potential Temperature (TO), Salinity (SO), and CO2 forcing concentrations.* **Projections:** Generates AMOC anomaly predictions for multiple lead times, including 15, 30, 45, and 60 years.* **Reproducibility:** Includes demonstration scripts (Python) and visualization tools (MATLAB) to verify the emulator's performance and reproduce key analytical figures.