Abstract: As a potential seismic noise source, train-induced seismic sources feature strong energy and can be utilized for near-surface passive surface wave imaging. However, train signals are highly directional, repetitive and non-stationary, which violates the assumption of a uniform source distribution in seismic interferometry. This leads to crosstalk during seismic interferometric processing and reduces the signal-to-noise ratio (SNR) of virtual shot gathers. Cross-coherence interferometry is applicable to sources with inherent coherence and delivers superior performance when processing coherent noise-source signals such as traffic-induced seismic noise. In this study, the processing workflow for train-induced seismic data is improved by introducing cross-coherence interferometry into seismic interferometry imaging using train noise. A comparative analysis is conducted between cross-correlation interferometry and cross-coherence interferometry in terms of imaging performance. The feasibility and effectiveness of the proposed method are verified using both synthetic and field data. High-quality virtual shot gathers are  obtained to extract dispersion curves, which are subsequently adopted for inversion. Finally, subsurface velocity profiles are constructed, achieving near-surface surface wave dispersion imaging with train-induced seismic sources. The results demonstrate that cross-coherence interferometry can effectively suppress crosstalk and markedly improve the SNR of virtual shot gathers when processing train-induced seismic signals.

Key words: train seismic signals, seismic interferometry, passive surface wave imaging, dispersion curve inversion