Advanced 3D Inversion of Passive Seismic Data for High-Resolution Subsurface Velocity Models in Tectonically Active Regions

Project Abstract

This study presents a novel framework for 3D inversion of passive seismic data to construct high-resolution subsurface velocity models in tectonically active regions, addressing key challenges of velocity heterogeneity, irregular station geometry, and limited sampling. We integrate ambient seismic noise cross-correlation (NCC) methods with robust full-waveform inversion (FWI) and a hierarchical Bayesian approach to jointly estimate shear and compressional velocity fields, attenuation, and anisotropy. The methodology begins with data preprocessing that emphasizes coherent signal extraction from long-term ambient noise using spectral whitening, cross-correlated function stacking, and time–frequency phase alignment to maximize signal-to-noise ratios while mitigating site effects. We then apply a multi-scale inversion strategy starting from 2D local earthquake and ambient noise-derived preliminary velocity models, progressively updating to full 3D anisotropic velocity structures through successive iterations of Green’s function computation, adjoint-state gradients, and line-search optimized step lengths. To manage non-uniqueness and sharp discontinuities typical of fault zones, we incorporate total-variation regularization and sparsity-promoting priors within a Bayesian sampler, enabling realistic sharp contrasts at fault interfaces and near-surface lithologic boundaries. The forward model couples 3D elastic wave propagation with frequency-dependent Q, capturing attenuation variations associated with fractured rocks and fluid-filled porosity, which are particularly pronounced in tectonically active environments. The inversion framework leverages GPU-accelerated solvers for scalable computation across dense receiver arrays and large model domains, enabling near-real-time iterations for iterative improvement. We validate the approach using a synthetic benchmark that emulates a complex faulted crust with heterogeneous velocity and anisotropy, followed by an application to an active fault zone with dense passive seismic deployments and supplementary borehole data. Our results demonstrate substantial improvements in resolving lateral velocity contrasts, near-surface velocity gradients, and anisotropic parameters relative to conventional NCC-only or isotropic inversions. Quantitative assessments show reductions in misfit by up to 40–60% and enhanced resolution of fault-associated velocity discontinuities and fracture zones. Uncertainty quantification provided by the hierarchical Bayesian framework yields credible intervals for each parameter, highlighting regions where data coverage limits resolution and informing targeted deployment of additional sensors. The study also analyzes sensitivity kernels to identify dominant data contributions from specific station pairs and frequency bands, guiding optimized survey designs for future monitoring campaigns. By delivering high-fidelity 3D velocity models with robust estimates of attenuation and anisotropy, this work improves subsurface characterization crucial for tectonic hazard assessment, mineral and hydrocarbon exploration, and geothermal reservoir monitoring. The proposed framework is adaptable to various regional settings, scalable to large datasets, and capable of integrating supplementary geophysical constraints such as gravity, magnetotellurics, and borehole measurements to further constrain the subsurface interpretation in complex tectonic environments.

Project Overview

What This Project Is About

---

The project explores how scientists use natural, passive seismic signals (like tiny vibrations from earthquakes or ocean waves) to create detailed images of what lies underground. It focuses on building a 3D method to infer how fast seismic waves travel through rocks, which helps map features like faults and fluids in tectonically active areas.

---

---

The Problem It Addresses

---

---

---

Objectives of the Project

---

1. Understand how ambient seismic signals reflect underground properties.
2. Develop a 3D inversion workflow to turn passive data into velocity models.
3. Test methods on synthetic (computer-made) data before real data.
4. Assess how well the models resolve important features like faults and fluid-filled rocks.
5. Evaluate uncertainties and robustness of the results.

---

---

What You Will Do Step by Step

---

1) Learn basics of seismology and inversion concepts. 2) Gather or simulate passive seismic data. 3) Build a 3D inversion framework. 4) Validate with synthetic datasets. 5) Apply to a real tectonically active area. 6) Compare with existing models. 7) Analyze uncertainties and sensitivity. 8) Document methods and results for reporting.

---

---

Expected Outcome

Deliverables include a functioning 3D inversion workflow, a set of high-resolution subsurface velocity models, and an evaluation of model reliability. The work could inform hazard assessment and resource exploration by improving underground maps in active regions.