*2.3. Postprocessing*

In the third step, a simple level-dependent wavelet threshold was applied to the signal obtained in the thresholding steps; the initial estimation of the seismic signal was subjected to CWT, shown in Equation (1); and thresholds of the coefficients of all scales were selected step by step using the hard threshold rule again, as shown in Equation (5). However, the threshold λ estimation used here was given by Donoho et al. [29], which was calculated using Equation (9): √

$$
\lambda = \sigma\_{\text{fl}} \sqrt{2 \ln N} \,\tag{9}
$$

where *σ<sup>n</sup>* = *median*- *Wy* /0.6745. Finally, the signal in the original data was extracted by applying an inverse CWT, as shown in Equation (10):

$$y(t) = \frac{1}{\mathbb{C}\_{\psi}} \iint \mathcal{W}\_{\mathcal{Y}}(a, \tau) d\tau \frac{da}{a^2}.\tag{10}$$

For the passive seismic raw records, the extracted signal *y*(*t*) represented the records that contained unnecessary prominent noise events, and the uniform background noise could be extracted by subtracting *y*(*t*) from the original data *y*.
