Three-dimensional First-arrival Traveltime Tomography

The first-arrival traveltime tomography is a crucial technique in land seismic data processing as it can assess the low wavenumber component of near-surface velocity structure,which could then be used in the static correction,the prestack depth migration and even the waveform inversion.With the prevalence of high-precision seismic exploration,the seismic data has gradually stepped into the era of big data.However,unlike the case of 2D,traditional traveltime tomography method could not meet the requirements of both computational accuracy and efficiency in 3D seismic data processing.Therefore,this paper focus on seeking the valuable improvement in the traditional tomography algorithm,which could enhance the practicability of tomography in dealing with large-scale 3D seismic data.First of all,the accurate first-arrival traveltime is the foundation of whole tomography technique.To ensure the precision and efficiency in 3D data processing,this paper builds up a new flow of picking based on the modified energy-ratio method.By utilizing the edge preserved smoothing operator and mispick auto-correction technique,this method not only enhances its applicability in seismic data with low signal to noise ratio,but also maintains the inherent high computational efficiency of energy-ratio method.The low accuracy and efficiency of traditional first-arrival traveltime tomography partially owe to the troublesome 3D ray tracing technique.Hence in this paper,an improved version of 3D fast marching method is proposed,which is called 3D multi-stencils fast marching method.It creates six stencils rather than one to solve the eikonal equation by rotating coordinate system.Besides,the implementation scheme of MSFM for irregularly topography and the associated posterior ray tracing algorithm based on Runge-Kutta method are also illustrated.Moving to the construction of tomographic equation,this paper introduces the basic concept of Fréchet derivative and establishes the multi-constraint objective function by combing the traveltime information,apparent slowness and model regularization,by which both the accuracy of inversion and the speed of convergence could be improved considerably.Furthermore,this paper also puts forward an efficient first-arrival traveltime tomography scheme based on the linearized eikonal equation to avoid the shortcomings inherited in the traditional ray-based tomography.The relationship between the traveltime perturbation and slowness perturbation is first set up by using the linearized eikonal equation.It is then solved as linearized inversion by employing the upwind finite-differences scheme and a conjugate gradient algorithm.Moreover,the shaping regularization method is also embedded in the CG algorithm to reduce the undetermined nature of inversion equation systems while enhancing the iterative convergence speed.Apart from the theoretical research of the first-arrival traveltime tomography technique,this paper also discusses its application in real seismic data,and illustrates the basic principles in choosing several core factors of seismic tomography,including offset,initial velocity,grid spacing,number of iterations,as well as the bottom elevation of the model.Besides,the evaluation criterion of inverted model is also introduced and the effectiveness and validity of 3D traveltime tomography algorithm are finally testified on the real 3D seismic data.In the last chapter of this paper,innovations and deficiencies of my work are concluded,and some expectations are bring forward.