*4.1. Preprocessing for Raw Field Data*

Raw field data will inevitably be affected by prominent noise events (e.g., mechanical construction), which makes it impossible to satisfy the assumption of a diffuse wavefield. Therefore, in order to reconstruct passive surface waves better, we first used the noise or signal removal algorithm based on SS-CWT proposed by Mousavi et al. [25] to preprocess the field data and separate the prominent noise events. Figure 3a shows an original field record of 90 s on which several relatively prominent noise events (red arrows) are

distributed. After noise and signal separation, the amplitude of the separated noise records was relatively evenly distributed (as shown in Figure 3b). Figure 3c shows the removed records containing prominent noise events (red arrows).

**Figure 3.** (**a**) Raw field data of 90 s. (**b**) Separated noise records. (**c**) Removed records. The red arrows indicate prominent noise events.

Power spectral density (PSD) plots are useful for visualizing variations in the frequency content of ambient data over time [33]. In order to further show the application effect of the noise separation algorithm in preprocessing the field data, we also calculated the PSD plots (as shown in Figure 4) corresponding to the three noise records in Figure 3.

**Figure 4.** PSD plots. (**a**) Raw field data. (**b**) Separated noise records. (**c**) Removed records. The white dotted boxes indicate prominent noise events.

Figure 4a shows the PSD plots of the original field data, on which a prominent highenergy event (white dotted box) can be seen. After noise separation, as shown in Figure 4b, the high-energy event in the white dotted box was attenuated, and the PSD of the whole record was evenly distributed. Figure 4c shows the PSD of the removed noise record, which was occupied by three prominent high-energy events (white dashed boxes). According to the results of Figures 3 and 4, the noise or signal removal algorithm based on SS-CWT could effectively remove the prominent noise events in the original passive seismic data.

### *4.2. Reconstruction of Virtual Shot Gathers by Cross-Correlation*

After preprocessing the raw field data, one geophone served as a virtual source for waves recorded by other receivers when using the cross-correlation calculation; thus, we could obtain all virtual shot gathers from every receiver without using an active source. In this paper, we showed the virtual shot gathered with the first receiver as the virtual source. In addition, in order to better characterize the intensity of the surface waves on the reconstructed virtual shot gathers, we used the dispersion curve image obtained by the phase-shift measurement method as the visualization tool.

Figure 5a shows the virtual shot gather formed by the cross-correlation of raw passive seismic data acquired in Sichuan for a certain hour. The surface waves (indicated by red arrows) can be seen in the red box on the virtual shot gather but are concealed by the energy of a strong transverse axis. The corresponding dispersion curve image is shown in Figure 5b, and the dispersion curve (red part) is disordered and discontinuous.

**Figure 5.** (**a**) Virtual shot gather formed by cross-correlation of raw one-hour field data and (**b**) the corresponding dispersion curve image.

After preprocessing the raw field data, the virtual shot gather formed by the separated noise records is shown in Figure 6a. Compared with Figure 5a, the in-phase axis (indicated by the red arrow) of the internal surface waves in the red box for the processed virtual shot gather is clearer, but it is still concealed by the energy of the strong horizontal axis. The corresponding dispersion curve image is shown in Figure 6b, where the curve (red part) is clearer and more continuous than in Figure 5b.

By comparing the virtual shot gathers and their corresponding dispersion curve images in Figures 5 and 6, we saw that the virtual shot gather formed by the separated noises had clearer surface wave in-phase axes and dispersion curves, which proved that the noise or signal removal algorithm based on SS-CWT could effectively enhance the reconstructed surface waves after preprocessing the raw field data.

**Figure 6.** (**a**) Virtual shot gather formed by cross-correlation of preprocessed separated noises and (**b**) the corresponding dispersion curve image.
