Study On Stable And Efficient Q Extraction And Inverse Q Filtering

In practice seismic waves traveling through the earth experience energydissipation and waveform distortion because of the anelasticity of the medium, whichcauses the S/N ratio and resolution of seismic data hardly to meet the requirements ofoil and gas exploration. The effective high frequency energy components in thePS-wave data are weaker that in the PP-wave data since rocks exhibit strongerabsorption for S-waves that for P-waves. In order to make better use of theinformation of P-and S-waves,we need to compensate for both phase and amplitudeof PP-and PS waves. Before inverse Q filtering, we need to accurately estimate Qvalues of P-and S-waves. Currently, however, the better methods of Q extraction andinverse Q filtering cost more time, especially for poststack data with large amount ofinformation. The low efficiency restricts the practical application of these methodsdue to high cost. On the purpose of saving time, it is necessary to improve thesemethods. Based on this idea, this paper mainly focuses on the research of the stableand efficient Q extraction and inverse Q filtering.The key to the inverse Q filtering compensation method is estimating relativelyaccurate P-and S-wave Q values. Generally, people adopt dimensionless Q factor toquantitively describe stratigraphic absorption and attenuation. So as to better simulatethe attenuation and dispersion, this paper studies the mechanism of seismic absorptionand attenuation and poses a contrast between the Kolsky-Futterman model, which iswidely used, and other models concerning the relationship between the attenuation,phase velocity, Q factor and frequency. The result indicates that the Cole-Cole modeland the SLS model differ a lot from other models at zero and infinite frequency;different models except the Cole-Cole model and the SLS model yield similar resultsin the seismic frequency range if the model parameters are available. When thefrequency is low, the SLS model and the Cole-Cole model show increasing Q fordecreasing frequencies. Azimi’s second model, Azimi’s third model, the power lawmodel, and Müller’s model show increasing Q for increasing frequencies.Kjartansson’s model is a constant Q-model because of frequency independent Q.Kolsky-Futterman model is a nearly constant Q-model because Q almost does notchange with frequency. The different models yield similar Q-values around the center frequency of the signal.Considering the development of Q extraction, this paper briefly introduces thebasic principles of three conventional Q estimation methods. According to the lowefficiency, this paper improves the spectral correlation-coefficient method. We add avarying window on each seismic event to avoid the influence of adjacent layers. Thesize of window depends on the traveltime difference between the current layer and theupper layer or lower layer. Then we transform the signals in the window intofrequency domain in order to compute the spectral correlation-coefficient. The Qaccuracy of different attenuation models demonstrates that this improved spectralcorrelation-coefficient method is faster than the spectral correlation-coefficientmethod, and has the same accuracy. Finally, the effect of inverse Q filtering of realdata shows that the improved spectral correlation-coefficient method is valid. To sumup, the improved spectral correlation-coefficient method is suitable for most seismicrecords attributing to the features of stronger anti-noise, higher efficiency and betterstability.Assuming that the subsurface layers are horizontal, we simulate the PS-waveray paths by ray tracing through each layer and compute the up-and down-goingpropagation times for each receiver time point based on the Snell law. Then we usecubic spline function to resample at each time point in the original records. Finally,we obtain all the travel times in each layer corresponding to each time point in thePS-wave records. Due to the asymmetry of the converted wave paths, we could notmake use of conventional P-wave methods to estimate S-wave Q values. This paperderives a method to estimate the S-wave Q values from common source gathers ofprestack converted wave and verifies the accuracy of this method. Besides, theaccuracy of P-wave will affect that of S-wave while using PS-wave to computeS-wave values.In order to make better use of the information of P-and S-waves,we need tocompensate for both phase and amplitude of PP-and PS waves to improve theresolution of seismic records by inverse Q filtering. This paper implements severalconventional inverse Q filtering methods and applies these methods to the syntheticdata of different layered Q models. We conclude that the layered inverse Q filteringmethod stably and efficiently compensates for seismic signals in deep layers and lowQ models. This method could recover all frequency components that in principle arerecoverable and could successfully suppress the artifacts.This paper extends the stable and efficient poststack inverse Q filtering algorithm to prestack inverse Q filtering for PP-and PS-waves, then improves the prestackinverse Q filtering algorithm to compensate for the seismic signals along thepropagation paths of PP-and PS-waves. Lastly, within the current constant Q layer,we decouple the amplitude operator which is a2-D function of traveltime andfrequency to the product of two1-D functions depending, respectively,on time andfrequency. The result of simulation shows that this algorithm obviously increases theefficiency of compensation on the premise of stability.