Soil Organic Carbon Simulation with Physics-informed Machine Learning (PIML) and Wavelet Decomposition

This release contains the scripts used to simulate and analyse soil organic carbon (SOC) trajectories under SSP climate scenarios using a physics-informed machine learning framework and Daubechies-4 wavelet decomposition.

The repository includes:

• PIML modelling scripts based on XGBoost with monotonic constraints • SOC projection under SSP126, SSP245, SSP370, and SSP585 • Ensemble and quantile uncertainty estimation • SHAP explainability analysis • Wavelet-based SOC trend decomposition • Regime-wise trend and sequestration rate analysis