Relating Megathrust Seismogenic Behavior and Subduction Parameters via Machine Learning at Global Scale

Abstract We investigate the relationship between the seismogenic behavior of global megathrusts and various subduction parameters. We performed a parametric approach by implementing three decision tree‐based Machine Learning (ML) algorithms to predict the b‐value of the frequency‐magnitude relationship of seismicity as a non‐linear combination of subduction variables (subducting plate age and roughness, slab dip, convergence speed and azimuth, distance to closest ridge and plate boundary). Using the Shapley Additive exPlanations (SHAP) to interpret the ML results, we observe that plate age and subduction dip are the most influential variables. The results suggest that older, shallow‐dipping plates contribute to low b‐values, indicating higher megathrust stress. This pattern is attributed to the higher rigidity of older plates, increasing flexural strength, and generating a shallow penetration angle, increasing the frictional interplate area and intensifying the megathrust stress. These findings offer new insights into the non‐linear complexity of seismic behavior at global scale.