In the denoising of seismic prospecting, background noise is often assumed to be stationary and Gaussian. However, this is not always appropriate for real seismic data. We used statistical tests to assess the stationarity and Gaussianity of land seismic data. The data we used for the analyses were passive noise records collected with receiver arrays in different land environments, e.g., deserts, steppes, and mountains. The results showed that the background noise was not strictly stationary, but locally stationary. The noise could be treated as a stationary series only in short time periods, whereas the stationarity became poor with increasing the time length of the noise records. By analyzing the behavior of noise data, we determined that the nonstationary noise always had more energy in the high-frequency band, which varied with the acquisition environments. The wind strength and the complexity of the environmental conditions also impacted the noise stationarity. Moreover, we found that noise in complex environments has a higher degree of Gaussianity than noise in simpler environment. Most of the noise records appeared to be a Gaussian process when examined over a period longer than 20 s, but they were found to be non-Gaussian when examined over shorter time periods of the order of 1 s.