High-performance CPU-GPU heterogeneous computing method for 9-component ambient noise cross-correlation

Abstract

Ambient noise tomography is an established technique in seismology, where calculating single- or nine-component noise cross-correlation functions (NCFs) is a fundamental first step. In this study, we introduced a novel CPU-GPU heterogeneous computing framework designed to significantly enhance the efficiency of computing 9-component NCFs from seismic ambient noise data. This framework not only accelerated the computational process by leveraging the Compute Unified Device Architecture (CUDA) but also improved the signal-to-noise ratio (SNR) through innovative stacking techniques, such as time-frequency domain phase-weighted stacking (tf-PWS). We validated the program using multiple datasets, confirming its superior computation speed, improved reliability, and higher signal-to-noise ratios for NCFs. Our comprehensive study provides detailed insights into optimizing the computational processes for noise cross-correlation functions, thereby enhancing the precision and efficiency of ambient noise imaging.