Improving the Retrieval of Empirical Green’s Function Using Adaptive Covariance Filter and Weighted Stacking

Seismic interferometry (SI) is widely utilized in fields such as seismic imaging and monitoring. The conventional approach involves segmenting long-duration ambient noise, performing cross correlation, and then stacking. However, when the sampling duration is short or the noise sources are unevenly distributed, the signal-to-noise ratio (SNR) of the reconstructed empirical Green’s functions (EGFs) is often low, hindering the precise extraction of dispersion curves. A common method to improve the SNR of EGFs is through weighted stacking strategies, yet these methods often face issues such as suboptimal improvement and limited applicability. In this article, we propose a novel weighted stacking strategy that incorporates the adaptive covariance filter (ACF) to enhance the reconstruction of EGFs. This method utilizes both time-domain similarity and phase consistency to improve the SNR of noise cross-correlation functions (NCFs). We conducted a comparative analysis of three SI methods—cross correlation, cross-coherence, and deconvolution. The analysis compared traditional linear stacking, standalone ACF, phase-weighted stacking (PWS), and time-frequency PWS (tf-PWS), as well as the combined effect of ACF with weighted stacking on the retrieval of Green’s function. The results indicate that the ACF-based weighted stacking method significantly improves the SNR of EGFs. Notably, the combination of ACF with tf-PWS yielded the most satisfactory retrieval outcomes. This study presents a promising approach for enhancing the accuracy and reliability of SI across various applications.