Improved adjoint-state traveltime tomography based on the eikonal equation: method, validation and application

Meng, Lingkai ; Lan, Haiqiang ; Badal, José (Universidad de Zaragoza) ; Guo, Gaoshan

Resumen: Seismic traveltime tomography methodologies — commonly categorized into wave-equation, ray-based, and eikonal equation-based approaches — face significant limitations in regional passive-source applications. Wave-equation methods deliver high-resolution imaging but are hindered by prohibitive computational costs and strong sensitivity to the initial velocity model. In contrast, ray-based and eikonal equation-based methods leverage efficient traveltime-velocity inversion schemes, but still exhibit fundamental divergences in their implementation. Among them, eikonal equation-based adjoint-state traveltime tomography (ATT) offers higher computational efficiency and lower memory usage through matrix-free formulations. However, conventional ATT implementations still suffer from two major limitations: (1) local extremes in the gradient, typically manifested as high-amplitude, short-wavelength artifacts caused by uneven ray coverage, and (2) substantial computational burdens caused by the imbalance between the numbers of seismic sources and receivers. To address these issues, we propose a novel ATT methodology that incorporates three key innovations: (1) preconditioned adjoint-state inversion, (2) spatially adaptive regularization to mitigate artifacts induced by non-uniform ray distribution and accelerate convergence, and (3) the application of the reciprocity principle to significantly improve computational efficiency. Synthetic experiments show that the proposed method not only improves geometric fidelity and amplitude recovery, but also achieves a 74-fold speedup per iteration compared to conventional approaches. When applied to the Southern California plate boundary, our approach further proves its robustness by resolving geologically consistent structures and detecting strong variations along the strike of the San Jacinto Fault, features that remain poorly imaged using conventional methods.

Idioma: Inglés

DOI: 10.1007/s10950-025-10341-9

Año: 2025

Publicado en: JOURNAL OF SEISMOLOGY (2025), [17 pp.]
ISSN: 1383-4649

Tipo y forma: Article (PrePrint)

Área (Departamento): Área Física Teórica (Dpto. Física Teórica)