Integrating artificial intelligence and physics in surface-wave methods: From automated analysis to physically consistent inversion

Surface waves provide critical constraints on the Earth’s near-surface and crustal structure, yet conventional analysis remains hampered by subjective dispersion picking, nonlinear and non-unique inversions, and high computational cost. Recent advances in artificial intelligence (AI) have begun to reshape every stage of surface-wave analysis—from data preprocessing to inversion and interpretation. This review synthesizes progress in AI-enabled surface-wave methods, including automated dispersion-curve extraction, dispersion-curve inversion, spectrum- and waveform-based approaches, probabilistic and physics-guided frameworks, and explainable-AI techniques. We highlight how deep learning achieves orders-of-magnitude acceleration over traditional workflows while addressing long-standing limitations such as mode interference and dependence on initial models. A near-surface case study further illustrates AI-assisted interpretation of tomographic results for karst-cavity detection. By comparing network Jacobians derived through automatic differentiation with classical physical sensitivity kernels, we reveal that data-driven inversions often exhibit non-physical sensitivity patterns, emphasizing the need for physically consistent and interpretable AI models. We conclude that future research should focus on improving generalization across geological settings, establishing standardized benchmark datasets, and developing physically informed architectures that unify data-driven learning with geophysical principles. Through these efforts, AI-integrated surface-wave methods are poised to transform seismic imaging from an experience-driven to a data- and physics-driven discipline.