Research On Processing Techniques Of Deep Seismic Reflection Data And Its Applicat Ion In Bangong-Nujiang Suture

Deep seismic reflection technique is one of the most efficient method for deep structure detection,it can describe the fine structure of the lithosphere.Deep seismic reflection technique derives from conventional seismic reflection method,which is widely used in petroleum industry.Both of them are based on the same principle,but the detection target is diferent between the two method.For deeper detection depth,deep seismic reflection technique has its own characteristics,such as huge explosive energy,long receiving spread,sparse spatial sampling rate,long recording time,etc.These acquisition characteristics lead to challenges and difficulties when processing the deep siesmic data.For example,huge explorsive energy and poor shooting-receiving conditions will increase the influence of groundroll wave,sparse spatial sampling rate will lead to space aliasing.It is nessesary for us to change the seismic data processing strategy to adapt the new challenges in deep seismic reflection data processing.In this thesis,I analyzed and summarized the characteristics of deep seismic reflection data and processing dificulties,and then 1 proposed a processing strategy aimed at deep seismic reflection data processing.To enhance the stacking result,I studied a comprehensive weighted stacking method.To depicts weak signal in deep and analysis the time-frequency characteristic of deep seismic reflection profile,I studied the time-frequency decomposition method based on matching pursuit.And I also studied parallel computing of the proposed method base on OpenMP to adapt the massive deep seismic reflection data.In the end of this thesis,I applied the proposed processing method into Bangong-Nujiang Suture(BNS)deep seismic reflection data processing.The specific research results and understandings mainly include four parts as listed below:?.Proposed a processing strategy based on the characteristic study of deep seismic reflection data.I.It is nessasery to test different static method when dealing with serious static problem caused by complicated near surface condition and drastic elevation fluctuations.Combine the low-frequency componet of datum statics and high-frequency componet of tomography statics(or refraction statics)to solve medium-long wavelength static problem.Then use frequency-divisional residual static method to solve short-medium wavelength static problem.2.Condidering the existence of space aliasing caused by sparse spatial sampling rate,f-k filter is avoid during deep seismic reflection data processing.Instead,the adaptive groundroll attenuation method is applied to remove ground roll,multiple processing window is applied to keep deep reflectors,which have simillar frequency and dip with the coherent noise at shallow depth,from being removed.3.To suppress random abnormal amplitude noise,multi-domain and frequency-divisional method is suggested.It is necessary to apply amplitude compensation in shot domain before multi-domain denoise process.4.Apply the velocity analysis method based on Anisotropy instead of conventional method based on hyperbola can correct the reflection on far offset more accurately.?.Studied a comprehensive weighted stack method to improve the stacking result of deep seismic reflection data.Traces in the prestack gathers are different,conventional stack method ignore the differences among traces in the prestack gather.As a result,some traces with low signal-noise ratio will decrease the quality of stacking result.In this study,I proposed a new stack method,which considers amplitude,frequeny,offset and incident angle factors to estimate the quality of each trace,and give different stack weights based on their quality during the stacking.This comprehensive weighted stack method will reduce the bad influence of those traces with low signal-nosie ratio.Apply this new stack method in deep seismic reflecion data will enhence the imaging quality,especially in the deep.?.Proposed a parallel algorithm for matching pursuit time-frequency decomposition method based on OpenMP.The matching pursuit method is based on Morlet wavelet,which can represent seismic signle more properly than Ricker wavelet.Morlet wavelets of seismic trace can be decomposed by matching pursuit method,then we can obtain a new profile reconfigurated by the decomposed Morlet wavelets within expected frequency range.By using this technique,weak sinal in the deep can be revealed,for example we can use this technique to depict the structure of Moho.Calculate the Smoothed Pseudo Wigner-Ville Distribution of each Morlet wavelets and stack them togather,we can get the time-frequency spectral of the original trace.In this way,we can obtain a high time-frequency resolution spectral without the influence of the cross-term.It is an effective method to analyze the time-frequency characteristic of deep seismic reflection profile.According to low-frequency shadowing effects,we can identify the potential partial-melting deep in the earth.In addtion,to improve the computation efficiency for massive deep seismic reflection data,a parallel algorithm based on OpenMP for this method is proposed.?.The proposed processing technique and strategy were applied in the deep seismic reflection data of BNS.The following findings and understandings are drawn based on the study of the deep seismic reflection profile.1.Deep seismic reflection profile revealed the exsistance of the collision process.2.Blank zone is widely distributed in the lower crust,which mean the lower crust may have experienced extensive remelting during the closing process of Bangong-Nujiang Ocean.Another explaination is that the plastic state is common state in the lower crust,thus the remelting is not necessary in this situation.3.The reflection feature of the inner Moho,which is depicted by wavelet reconfiguration technique,shows that the newborn Moho suffered S-N compression,which is probably related to the India-Eurasia collision.The process result shows that the Moho under BNS and southern Qiangtang terrane is relatively flat.The Moho is around 57-66km deep.4.The north-dipping brigth-spots,which located in the bottom of the lower crust of the northmost Lhasa terrane.The main features of the bright-spots are:negative polarity,strong amplitude.Low-frequency shadow is detected by utilizing the time-frequency technique based on matching pursuit.Partial-melting is very likely existed under the bright-spots,and the bright-spots represent the boundray between solid rock and molten mass.