The Study And Test On The Properties Of The Environmental Noise In The Land Seismic Exploration

Land seismic exploration is an important exploration in coal and oil exploration.But in the process of field acquisition, the seismic signal inevitably submerged in allkinds of random noise, which make the seismic data processing difficult and the sameto the interpretation of the subsequent data analysis, with a direct impact on theaccuracy of the exploration results. The study of seismic random noise has always beenfocused on the noise suppression hoping to have a powerful noise reductioncountermeasure.In the field of noise suppression, the traditional understanding to the properties ofseismic noise: linear, stationary, gaussian. It is desired of noise properties that areconvenient to provide theoretical basis for selecting seismic random noise suppressionmethod. However, in some cases, de-noising method based on the idea is not good. Itneeds a new understanding to the properties of random noise to challenge thetraditional understanding.As an important component of random noise, there is no scientific study on theproperties of environmental noise. The predecessors are always based on the intuitiveassumption. In this paper, using the unbiased estimation theory of Gaussian noise in thetime-frequency peak filter, I de-noise the actual record simulated seismic data whichrespectively mixed with the actual pure noise and the gaussian white noise. Comparethe SNR after de-noising, it believe that the actual noise may be non-gaussian. Usingthe same de-noising thought for the stationary process of wiener filter, there also thinkthe non-stationary of the noise existed. So it is necessary to test the stationarity andgaussian.This paper has quantitatively and qualitatively studies on the stationarity andgaussian of four areas as forest, desert, mountains and loess tableland. The method oftesting the stationarity is based on the empirical Fokker-Planck function which is anonparametric method. I consider a short interception of data points for what thecurrent filter algorithm generally used frequency domain filtering or for local data within a short time window filtering. Comparing the stationarity of the short-term noiseand analysis the variance, the results show that the measurement of the noises arenon-stationary, and the degrees of stationary are also different. So further to analysisthe local stationarity conditions of the noise, the noise can be local stationarity study.Normal test of the four common methods use the Monte Carlo experiments, whichinclude the S-W test, K-S test, Lilliefors tests and J-B test. The S-W test is the mostpowerful test as the actual noise testing method. The test results show that thenon-gaussian weakened in turn as the desert, loess tableland, mountain and forest. Butwhen the data length is long enough, the measuring result show that the actual seismicdata is entire non-gaussian.The paper mainly research and test the properties of environmental noise in theland seismic exploration, which includes stationarity and gaussian. Through systematicstudy on the different regions and the different lengths of pure noise record, theresupplement the studies of properties on the land seismic environmental noise,providing theoretical basis for modeling noise and de-noising according to theproperties.