关键词: 深地震反射数据, 稀疏约束反演谱分解, 时频相位谱, 深部细结构, 数据重构

Abstract:

Deep seismic reflection detection technology is considered to be one of the effective technical means to detect the fine lithospheric structure and the deep structure in earth’s interior. However, the detection depth of deep seismic reflection technology is large, and the geological tectonics of target area are usually complex(such as there are many fault zones, the terrain elevation changes greatly), which makes the explosion condition of source and the receiving condition of seismic wave very poor in the process of seismic data acquisition. This results in that the deep seismic reflection data has low frequency, narrow frequency band, low signal-to-noise ratio and resolution, and weak reflection energy. In addition, due to the existence of various forms of complex geological bodies in the deep crust and the energy absorption and attenuation caused by the long-distance propagation of seismic wave, etc., the reflection wave groups of the middle and lower crust in the deep seismic reflection profile is often characterized by weak reflection energy, poor continuity, interleaved or zonal distribution, which brings difficulties to the correct interpretation of seismic data and the understanding of geological structure.
With the emergence of high-resolution time-frequency analysis methods, it becomes more and more meaningful to try to use new time-frequency analysis methods to analyze the time-frequency distribution characteristics and phase extraction of deep seismic reflection data. The sparse constraint inverse spectral decomposition(SCISD)is a time-frequency analysis method based on the unsteady convolution model. This method describes the spectral decomposition problem as a linear inversion problem, in which minimizing the l1 norm of the time-frequency spectrum of the seismic signal is adopted as a sparsity constraint term, and it has the ability to generate high-resolution time-frequency spectrum and phase spectrum. Compared with transform-based spectral decomposition methods(such as short-time Fourier transform, continuous-wavelet transform and S-transform), this method has higher time-frequency resolution, more concentrated time-frequency distribution, and good anti-noise performance. It can extract the corresponding time-varying wavelet phase information, and is not affected by the shape and length of time window. Due to these good features, the SCISD method has achieved certain results in low-frequency anomaly detection, reservoir identification and carbonate cave-type reservoir prediction for petroleum geophysical exploration. On the one hand, it can provide high time resolution for thin reservoir prediction, on the other hand, it can provide wavelet phase change characteristics for gas reservoir detection and carbonate cave identification. However, it has not yet been applied to deep seismic reflection data. Therefore, this paper applies the SCISD method to the deep seismic reflection data in the Sanhe-Pinggu M8.0 earthquake area. The test of single-channel data shows that this method can effectively identify the seismic amplitude anomaly and reconstruct the seismic signal more accurately. The time-frequency characteristic analysis and data reconstruction of the deep seismic reflection data in the Sanhe-Pinggu M8.0 earthquake area were carried out. It is found that the reconstructed section and time-frequency section have higher signal-to-noise ratio and resolution than the original section, and improve the continuity of the seismic profile events. It can not only calibrate the frequency components at different times of weak signals in the deep seismic reflection data, but also clearly depict the characteristics of low-frequency details on the stacked section. The high-resolution time-frequency phase spectrum has rich phase information, and the phase characteristics change significantly. It can be used for fault identification, which is helpful to judge the tectonic boundary. The wavelet reconstruction technique and the high-resolution single-frequency phase spectrum are helpful to characterize the internal reflection structure of the Moho transition zone. The SCISD method is suitable as the spectral decomposition method for deep seismic reflection data.
By analyzing the time-frequency characteristics of the original reflection seismic data and purposefully reconstructing the deep seismic reflection data, the effective weak signals in the deep crust can be extracted. The joint comparative analysis of time-frequency profile and reconstructed profile is helpful to deeply study the deep fine structure and establish an intuitive understanding of the deep structure. This method analyzes the deep seismic reflection profile from a new perspective, which is beneficial to discover new geological understandings, has very important research significance, and can be widely used in the interpretation of deep seismic reflection data.

Key words: deep seismic reflection data, sparsity constraint inverse spectral decomposition, time-frequency phase spectrum, crustal fine structure, data reconstruction