The isotropy of ambient noise sources is a fundamental prerequisite for accurately obtaining Green’s functions, which are widely used in ambient noise imaging and seismic velocity monitoring. When noise sources are evenly distributed, sufficiently long term noise cross-correlation can be used to extract the empirical Green’s functions between stations, thereby obtaining subsurface medium information between station-pairs without relying on earthquake waveform records. However, due to the seasonal influence of marine activity, the azimuth of microseismic noise sources shifts, and the energy fluctuates. When the noise source is evenly distributed, the positive and negative half branches of the empirical Green’s function exhibit a symmetric state. However, the empirical Green’s function between station-pair often shows asymmetric characteristics, with different energy levels propagating in the two directions corresponding to the positive and negative half branches. Therefore, using the signal-to-noise ratio of the positive and negative half branches of the empirical Green’s function can reveal the azimuthal distribution of noise sources. The eastern part of Chinese mainland is bordered by the sea, while the western part faces land, with a complex internal geological structure that includes basins, plateaus, and mountain ranges. It also has a large number of high quality digital seismic stations, making it an ideal natural laboratory for studying the azimuthal characteristics of microseismic noise sources. Many scholars have studied the azimuthal characteristics of noise sources in different regions of Chinese mainland using various methods. However, these research results vary depending on the data used and the distribution of the recording stations, leading to uncertainty in the azimuthal distribution of dominant noise sources. Therefore, it is necessary to take Chinese mainland as a whole and use a unified dataset to comprehensively investigate the noise characteristics of Chinese mainland.

This study focuses on the asymmetric characteristics of the empirical Green’s function, using 377 770 annual records of inter-station empirical Green’s functions with periods ranging from 8 s to 50 s, published by Xiao et al. It primarily investigates the azimuthal characteristics of microseismic noise sources in Chinese mainland and provides a preliminary analysis of their excitation mechanisms. The study finds that the amplitudes of the signal-to-noise ratio within two periods show a spatial distribution that is related to the geological structure of Chinese mainland: the amplitudes are lower in the Qinghai-Xizang Plateau and higher at the boundaries where multiple tectonic blocks intersect. This spatial distribution is linked to the uneven subsurface velocity structure, as indicated by the shear wave velocity structure at a depth of 10−20 km. Additionally, the Bohai Bay Basin exhibits a significantly lower signal-to-noise ratio than the surrounding regions, which may be related to the thick, soft sedimentary deposits in the basin.

The microseismic noise sources in Chinese mainland show significant spatial heterogeneity: the azimuthal anisotropy and dominant azimuths of the second type of seismic tremor （8−10 s） have the highest signal-to-noise ratios along the coast, while the first type of tremor （10−20 s） shows the opposite pattern. These azimuthal distribution features reflect the excitation mechanisms of the noise sources: ① In the nearshore region, incident and reflected ocean waves interact with each other, exciting the second type of seismic tremor （8−10 s）. The dominant azimuths for this period mainly point to the Indian Ocean and the South Pacific, followed by the South Atlantic. ② Statistical analysis of the effective wave height of mixed seas over 30 years shows the regions with extreme values of effective wave height are primarily located in the South Indian Ocean, North Atlantic, and North Pacific. These areas, which have higher energy wave activities, show high consistency with the distribution of the first type of seismic tremor in this study. Moreover, deep sea wave movements are more likely to generate low frequency microseismic events, especially in mid-to-high latitude deep-sea regions such as the South Indian Ocean and North Atlantic. Based on this, it is inferred that the first type of seismic tremor （10−20 s） is excited by deep sea ocean waves interacting with the far sea seafloor, primarily originating from the North Atlantic, followed by the deep water regions of the Indian Ocean and North Pacific. It should be noted that the annual averaged data somewhat eliminates the seasonal variations of noise sources, but also loses seasonal characteristics. Therefore, the azimuthal distribution characteristics of the noise sources in this paper are annual averages and may have certain azimuthal deviations when compared to previous regional results. When the station distribution is dense and uniform, the cross-correlation data for each azimuth are sufficiently available, and the noise source azimuthal characteristics obtained will be more accurate. However, the actual station distribution is uneven, which may lead to missing cross-correlation functions in certain directions and potentially affect the results. The study of the azimuthal characteristics of microseismic noise sources not only helps to understand the relationship between oceanic motion and the excitation mechanisms of noise sources, but also provides important azimuthal prior information for ambient noise-based imaging techniques.