Statistical properties of the random noise in seismic data

In the denoising process of seismic prospecting, random noise is often assumed to be stationary and Gaussian. However, this is not always appropriate for real seismic data. In this study, we used the statistical testing methods to investigate the properties of land-seismic-prospecting random noise on stationarity, Gaussianity and linearity. The dataset was collected by the seismic arrays at different typical environments, e.g., desert, grassland and hilly areas. The results show that the random noise is a locally stationary series, and it can be treated as a stationary series only in short time periods. Moreover, the random noise cannot be considered as a Gaussian process, and the noise in a complex environment always has a better Gaussianity. The linearity of the random noise varied with the complexity of the acquisition environment, and in terms of linearity, the noise in a simple environment is superior to that in a complex environment.