Study Of Near-surface Seismic Velocity Inversion,Imaging,and Migration Velocity Analysis In Time Domain

Geophysical exploration methods image the underground structures according to different physical characteristics of rocks.Seismic exploration is one of the most important composition of geophysical exploration methods.According to the changes of wave propagation,vibration intensity and waveform characteristics when seismic wave encountering different elastic interfaces,we can interpret the geological structure.In the exploitation of oil and gas,mineral resources and other resources,seismic exploration method,as one of the most important exploration methods,has received more and more attention.With the increasing demand for energy,exploration operations are facing more and more complex near-surface structure,as well as higher imaging accuracy and efficiency requirements.In the near-surface area,the topography is obviously undulating,the structure is complex,and the lithology changes greatly,which brings great difficulties to data processing,deep structure imaging,reservoir prediction,and geological interpretation.High-resolution near-surface velocity model provides more accurate statics values for seismic data processing and lays the foundation for deeper velocity model building.Migration is one of the most important research directions in seismic exploration.Accurate migration velocity calculated by velocity analysis is an essential precondition of prestack time migration imaging.This paper focuses on the near-surface velocity inversion,imaging,and subsurface velocity analysis in seismic exploration:1.Refraction traveltime migrationRefraction traveltime migration is an efficient and accurate method for underground refractor imaging.Refraction traveltime and overburden velocity information are used for calculation.This method can be used to solve the problems of low resolution and unclear layer structure in seismic inversion.The method involves the reconstruction of the forward and backward wavefront in the subsurface by applying a wavefront raytracing method.Subsequently we apply the Hagedoorn imaging condition to generate the precise shape of the high-velocity refractor.We analyze the uncertainties of this method based on one synthetic model.For a real dataset,the refraction traveltime migration provides a more accurate long-wavelength statics value,which helps to improve the common midpoint(CMP)stacking quality.We further extend the refraction traveltime migration method to 3D.According to the actual 3D geometry,a new imaging condition is developed.Based on the 3D method,we can image 3D refractors with high accuracy and efficiency.2.Joint inversion of refraction traveltime tomography and migrationAccurate refractor image is of great significance to improve the resolution of velocity model.In order to stabilize the inverse problem in velocity model building,regularization term is usually added in the objective function.Therefore,traditional inversion methods tend to produce a smooth velocity model.If the true shallow structures include a weathering layer over a high-velocity bedrock,traditional velocity model building methods,such as first-arrival traveltime tomography,may fail to image the sharp interface with high resolution.We intend to obtain a high-resolution velocity model with the help of refraction traveltime migration method,which could image the refractor with high accuracy.In this study,we propose a joint inversion method of refraction traveltime tomography and migration.The migration image is applied to constrain traveltime tomographic inversion by imposing discontinuities at the refraction interfaces in model regularization.In each subsequent iteration,the shape of the migrated refractors and the velocity model are simultaneously updated.In order to adapt to more complex situations,we further extend the refraction traveltime migration method to multi-layer structure,which can accurately image all layers and velocities.3.Study and application of elastic waveform inversion methodsThe traveltime imaging methods can effectively establish a relatively macro velocity model.For more precise structures,such as low-velocity layers and small scatters,it is necessary to use the waveform inversion methods to further improve the model resolution.Elastic full waveform inversion(FWI)is a popular data fitting method used to estimate seismic properties of the earth.It uses all seismic wave information including amplitude,phase,and frequency to produce high-precision imaging of underground structures.FWI requires initial models with high accuracy,especially in the complex area.Meanwhile,when the seismic data is lack of low frequency component,it is very easy for FWI to occur cycle skipping.We present the field seismic data collected at Mount St.Helens(MSH),aiming at imaging the pyroclastic density produced by 18 May 1980 volcano eruption using elastic full waveform inversion with high resolution.The near-surface structure of this area is complex and lack of low frequency data,which brings great challenges to the application of FWI.We develop a specific workflow to improve the resolution of velocity models step by step.The initial compressional wave speed(Vp)model is constructed using data from traveltime to early-arrival waveform,and initial shear wave speed(Vs)model is imaged by Multi-channel Analysis of Surface Waves(MASW)method.For the inversion process,we start with envelope inversion and then move on to waveform inversion.In both Vp and Vs models,the feature structures in final inverted models are quite similar with the adjacent outcrop,which indicate the accuracy of these imaging results.4.Prestack time migration velocity pickingAccurate near-surface velocities provide the foundation for deeper velocity model building.Migration helps to move inclined layer to a correct position,which can improve the convergence of diffractive wave and homing of turning wave.The accuracy of migration velocity directly determines the quality of the migration image.How to get an accurate migration velocity is a key problem in seismic exploration.Traditional migration velocity analysis includes both velocity spectrum generation and interactive manually velocity picking.This step usually takes a lot of time,and requires high level of picking quality,which is greatly affected by human.In this study,we develop an automatic prestack time migration velocity picking method based on U-Net.In the neural network training process,the prestack time migration profile migrated by constant velocity is treated as input,root mean square(RMS)velocity model is taken as output.Migration velocity with high precision can be obtained by small amount of learning samples.Field seismic data test indicates the effectiveness of this method.