A Predictive Framework for Multi-Scale Seismogenic Hazard Coupling

Thesis Abstract

This study addresses the growing need for predictive capabilities in seismogenic hazard by integrating multi-scale processes—from fault-scale dynamics to regional tectonic stress fields—to develop a coherent framework for hazard coupling and forecasting. The problem centers on fragmentation between disciplinary scales that impedes timely risk assessment and mitigation planning in seismically active regions. The aim is to formulate and validate a predictive framework that links microseismicity, fault creep, and lithospheric stress accumulation to probabilistic hazard estimates across spatial scales. Specific objectives include (i) characterizing the interactions between fault-zone frictional properties and background seismicity using high-resolution earthquake catalogs; (ii) synthesizing geodetic time-series (GNSS, InSAR) with seismicity data to quantify scale-dependent coupling factors; (iii) developing a multi-scale stochastic model that integrates near-field rupture processes with regional seismicity patterns; (iv) validating the framework against retrospective earthquake sequences and prospective forecasts in a major seismogenic belt; and (v) delivering decision-ready hazard indicators for emergency management and infrastructure resilience. The methodology adopts an explanatory sequential mixed-methods design. The population comprises global high-seismicity regions with dense data coverage, with a focused case study in the circum-Pacific belt. The sample includes earthquake catalogs spanning the past five decades (n > 120,000 events with M ? 3.0), continuous GNSS time series (15–20 stations per site, multi-year records), and Interferometric Synthetic Aperture Radar (InSAR) displacement maps for selected fault zones. Data collection instruments encompass publicly available seismic catalogs (USGS, IRIS), geodetic datasets (IGEOD, EUREF Permanent Network), and satellite-derived deformation products from Sentinel-1 and ALOS-2 missions. Validity and reliability are ensured through cross-validation of catalog magnitudes, declustering for independent seismicity, quality filtering of GNSS time series, and calibration of InSAR deformations with borehole strainmeters where available. The analytical framework integrates (i) probabilistic seismology to quantify scale-dependent hazard using time-dependent Poisson and non-Poisson renewal processes; (ii) hierarchical Bayesian modeling to fuse fault-scale friction parameters with regional tectonic loading indicators; (iii) multi-variate regression and machine learning approaches (random forests and gradient boosting) to identify key predictors of seismic bursts across scales; (iv) couplets of physics-based fault friction models (pring, rate-and-state friction) with stochastic renewal models to generate synthetic multi-scale hazard scenarios; (v) network analysis to examine spatial-temporal clustering and cascade effects; and (vi) sensitivity analyses to evaluate uncertainties in data inputs and model structure. Model specification includes a hierarchical state-space representation that links microseismic parameters to meso- and macro-scale hazard indicators, with calibration performed via Markov Chain Monte Carlo (MCMC) sampling and Bayesian model averaging to compare competing coupling hypotheses. The study also implements an ethical data governance plan for handling sensitive geohazard information. Expected findings indicate that incorporating multi-scale coupling improves forecast skill for short-term seismicity patterns (7–90 day horizons) while preserving robust long-term hazard estimates at regional scales. The research anticipates identifying specific coupling regimes where near-field frictional instability signals propagate into regional seismicity, manifested as statistically significant lead-lag relationships between geodetic velocity anomalies and subsequent earthquake sequences. It is expected that the framework will reveal heterogeneous coupling strength across different tectonic settings, with fault-zone properties modulating the transmission of stress perturbations to broader seismic networks. The contribution to knowledge lies in operationalizing a unified, scalable predictive framework that bridges microphysical fault processes and macro-scale seismic hazard, validated by retrospective case studies and prospective testing. The main conclusion is that a multi-scale, probabilistic coupling framework enhances predictive capability for seismogenic hazards and supports evidence-based risk mitigation. Recommendations include integrating the framework into regional seismic early warning systems, expanding geodetic networks in obscured fault zones, and committing to open-data sharing and continuous model updating as new geophysical observations become available.