Comparative Analysis of Seismic Hazard across Subduction Zones Globally

Thesis Abstract

The study addresses the uneven distribution of seismic hazard across global subduction zones and the limitations of current hazard models to capture local-to-regional variability in shaking intensity, rupture behavior, and tectonic regime transitions. The aim is to develop a comparative, cross-regional assessment of seismic hazard by integrating physico-tec tonic, statistical, and stochastic modelling approaches to produce harmonized hazard estimates and identify drivers of regional differences. Specific objectives are to (i) compile a globally representative database of subduction-zone earthquakes (magnitude Mw 5.5–9.5) from 1900–2023, (ii) evaluate differences in ground-motion prediction models (GMPs) performance across subduction segments using residual analysis and information criteria, (iii) quantify probabilistic seismic hazard using uniform logic-tree weighting and site amplification scenarios for five representative subduction zones (Andean, Cascadia, Aleutian, Circum-Pacific, and Sunda) and their distinct asperity patterns, (iv) investigate the role of slab geometry, interface coupling, and slab window dynamics on hazard metrics via mixed-effects regression and hierarchical Bayesian models, and (v) synthesize results into regionally harmonized hazard curves and a decision-support framework for infrastructure design and emergency preparedness. A mixed-methods research design is employed, combining quantitative seismotectonic analysis with probabilistic seismic hazard assessment (PSHA) and scenario-based planning. The population comprises globally distributed subduction-zone seismic events and corresponding ground-motion records from accelerometer and broadband seismograph networks. The sample includes all documented significant events (Mw ? 5.5) within twenty subduction segments, yielding approximately 2,850 records and 320 events after screening for depth, mechanism, and data completeness. Data collection instruments include high-resolution earthquake catalogs (USGS, INGV, JMA), seismic moment tensors, moment magnitudes, rupture dimensions, site-condition databases (Vs30, soil classes), and GMPE archives. Analytical techniques encompass (i) regression-based evaluation of GMPE performance with multivariate residual diagnostics, (ii) hierarchical Bayesian PSHA incorporating logic-tree uncertainty and regional damping factors, (iii) stochastic ground-motion simulation using the modal summation and inverse fast Fourier transform methods for scenario sets, (iv) sensitivity analyses of slip distribution, rupture speed, and frictional properties on hazard outputs, and (v) geodynamic modelling to relate slab geometry and coupling to observed hazard trends. The study applies theoretical lenses from global Seismo-Tectonic Theory and Vulnerability Theory, with explicit references to the Gutenberg–Richter law, the Seed weighting scheme for GMPE comparison, and the framework of Conditional Probability of Exceedance in PSHA. Validity and reliability are addressed via cross-validation of GMPE fits, bootstrapping of hazard curves, and inter-laboratory data checks. Expected findings include (i) identification of systematic GMPE biases by subduction-zone regime and site condition, (ii) quantification of variability in peak ground acceleration and spectral acceleration across zones, (iii) evidence that slab geometry and interface coupling significantly modulate hazard peaks and low-frequency response, and (iv) demonstration that harmonized hazard curves across regions improve consistency for infrastructure design in transitioning tectonic environments. The study anticipates that high-seismicity regions with complex asperity structures (e.g., Cascadia and Sunda) exhibit greater epistemic uncertainty in hazard estimates, whereas more uniform coupling regions (e.g., Circum-Pacific megathrusts) show relatively stable hazard profiles. Contribution to knowledge includes (i) a globally comparative, methodologically rigorous PSHA framework tailored for subduction zones, (ii) an integrated dataset linking geodynamics with hazard metrics suitable for researchers and practitioners, and (iii) practical decision-support insights for engineers and policymakers regarding site-specific design spectra and hazard mitigation planning. The main conclusion is that seismic hazard in subduction zones cannot be universally characterized by a single regional model; instead, hazard estimation benefits from regionally informed hyperparameters and explicit consideration of slab dynamics, which reduces epistemic uncertainty in threat assessments. Recommendations emphasize expanding high-fidelity ground-motion databases in underrepresented subduction zones, improving taxonomies for asperity distribution, adopting harmonized hazard products in building codes, and fostering interdisciplinary collaboration among geophysicists, engineers, and urban planners to translate hazard knowledge into resilient infrastructure and effective emergency response strategies.