A Comprehensive Review of Earthquake and Ground Motion Prediction Methods Across Various Tectonic Settings: Conventional and AI-Based Models

Abstract

Earthquake and ground motion prediction are fundamental challenges in seismology and considered essential for seismic hazard assessment and early warning applications. The diversity of tectonic settings which include convergent, divergent, transform, and intra-plate aspects has introduced distinct geophysical behaviors influencing the design, reliability, and transferability of models. Therefore, this study aims to present a comprehensive review of recent advances in conventional and artificial intelligence (AI)-based prediction methods published between 2019 and June 2025. The methodology adopted was systematic selection and synthesis of over 76 peer-reviewed studies conducted based on both physics-based frameworks and modern artificial intelligence methods. The analysis showed a progressive evolution from empirical and statistical formulations toward physics-informed deep learning models developed by integrating spatial–temporal correlations and physical constraints. Convergent and transform boundaries were observed to have dominated current studies due to higher seismic energy release and richer data availability while divergent and intra-plate zones remained underexplored. Meanwhile, hybrid frameworks and ensemble deep neural networks improved accuracy and generalization across tectonic contexts compared to standalone methods. However, challenges persist in model interpretability, cross-regional scalability, and integration into real-time Earthquake Early Warning Systems. The synthesis showed the need for closer collaboration between data intelligence and seismology towards providing globally adaptive and physics-informed predictive intelligence for future seismic risk mitigation.