Traveltime computation for 3D anisotropic media by a finite-difference perturbation method

The first-order perturbation theory is used for fast 3D computation of quasi-compressional (qP)-wave traveltimes in arbitrarily anisotropic media. For efficiency we implement the perturbation approach using a finite-difference (FD) eikonal solver. Traveltimes in the unperturbed reference medium are computed with an FD eikonal solver, while perturbed traveltimes are obtained by adding a traveltime correction to the traveltimes of the reference medium. The traveltime correction must be computed along the raypath in the reference medium. Since the raypath is not determined in FD eikonal solvers, we approximate rays by linear segments corresponding to the direction of the phase normal of plane wavefronts in each cell. An isotropic medium as a reference medium works well for weak anisotropy. Using a medium with ellipsoidal anisotropy as a background medium in the perturbation approach allows us to consider stronger anisotropy without losing computational speed. The traveltime computation in media with ellipsoidal anisotropy using an FD eikonal solver is fast and accurate. The relative error is below 0.5% for the models investigated in this study. Numerical examples show that the reference model with ellipsoidal anisotropy allows us to compute the traveltime for models with strong anisotropy with an improved accuracy compared with the isotropic reference medium.     

基于二维三次卷积插值算法的辛几何射线追踪

射线追踪是地震波走时层析成像的基础,射线空间位置的准确性及射线走时的精度决定了层析成像的可靠性.本文根据哈密尔顿系统可以有效提高程函方程解稳定性的特性,采用辛几何算法（SAM-Symplectic Algorithm Method）及二维三次卷积插值技术进行地震波射线追踪.由于采用了SAM算法,保证了地震波波前精度,提高了射线空间位置的准确性.数值模拟结果表明SAM既能保证哈密尔顿系统的稳定性又具有运算速度快的特点,提高了射线追踪的计算精度.     

An upwind fast sweeping scheme for calculating seismic wave first‐arrival travel times for models with an irregular free surface

The topography‐dependent eikonal equation formulated in a curvilinear coordinate system has recently been established and revealed as being effective in calculating first‐arrival travel times of seismic waves in an Earth model with an irregular free surface. The Lax–Friedrichs sweeping scheme, widely used in previous studies as for approximating the topography‐dependent eikonal equation viscosity solutions, is more dissipative and needs a much higher number of iterations to converge. Furthermore, the required number of iterations grows with the grid refinement and results in heavy computation in dense grids, which hampers the application of the Lax–Friedrichs sweeping scheme to seismic wave travel‐time calculation and high‐resolution imaging. In this paper, we introduce a new upwind fast sweeping solver by discretising the Legendre transform of the numerical Hamiltonian of the topography‐dependent eikonal equation using an explicit formula. The minimisation related to the Legendre transform in the sweeping scheme is solved analytically, which proved to be much more efficient than the Lax–Friedrichs algorithm in solving the topography‐dependent eikonal equation. Several numerical experiments demonstrate that the new upwind fast sweeping method converges and achieves much better accuracy after a finite number of iterations, independently of the mesh size, which makes it an efficient and robust tool for calculating travel times in the presence of a non‐flat free surface.     

SEISMIC REFLECTION AMPLITUDES1

Seismic amplitude variations with offset contain information about the elastic parameters. Prestack amplitude analysis seeks to extract this information by using the variations of the reflection coefficients as functions of angle of incidence. Normally, an approximate formula is used for the reflection coefficients, and variations with offset of the geometrical spreading and the anelastic attenuation are often ignored. Using angle of incidence as the dependent variable is also computationally inefficient since the data are recorded as a function of offset. Improved approximations have been derived for the elastic reflection and transmission coefficients, the geometrical spreading and the complex travel-time (including anelastic attenuation). For a 1 D medium, these approximations are combined to produce seismic reflection amplitudes (P-wave, S-wave or converted wave) as a Taylor series in the offset coordinate. The coefficients of the Taylor series are computed directly from the parameters of the medium, without using the ray parameter. For primary reflected P-waves, dynamic ray tracing has been used to compute the offset variations of the transmission coefficients, the reflection coefficient, the geometrical spreading and the anelastic attenuation. The offset variation of the transmission factor is small, while the variations in the geometrical spreading, absorption and reflection coefficient are all significant. The new approximations have been used for seismic modelling without ray tracing. The amplitude was approximated by a fourth-order polynomial in offset, the traveltime by the normal square-root approximation and the absorption factor by a similar expression. This approximate modelling was compared to dynamic ray tracing, and the results are the same for zero offset and very close for offsets less than the reflector depth.     

复杂介质中地震波前及射线追踪综述

本文较为系统综述了国内外在不均匀介质中各种主要和实用的射线追踪方法,例如:基于射线理论的打靶法、弯曲法(伪弯曲法)、高斯射线束算法等;基于网格单元扩展的有限差分解程函方程法(FD)、最短路径算法(SPM);以及结合射线和网格单元扩展的波前构造法等.同时对目前出现的多次波射线追踪技术、以及多值波前追踪技术(如:相空间算法、水平集算法)也进行了分析讨论.同时对基于网格单元扩展算法的优缺点进行了评述,其基本结论是:基于单元模型的SPM要优于FD算法,而基于网格的SPM算法则次之.就传统的射线追踪算法(如:打靶法和弯曲法)而言,其未来的发展方向是实现完全非线性的相应算法,而基于网格单元的算法则主要是扩展功能(如:后续波、多值波前的追踪).射线追踪方法技术未来需要解决的问题主要有:块状模型中多次波的追踪;多值波前及多值射线追踪;走时与振幅的同时追踪计算;以及其它领域新方法的引入.     

A new direction for shallow refraction seismology: integrating amplitudes and traveltimes with the refraction convolution section

The refraction convolution section (RCS) is a new method for imaging shallow seismic refraction data. It is a simple and efficient approach to full‐trace processing which generates a time cross‐section similar to the familiar reflection cross‐section. The RCS advances the interpretation of shallow seismic refraction data through the inclusion of time structure and amplitudes within a single presentation. The RCS is generated by the convolution of forward and reverse shot records. The convolution operation effectively adds the first‐arrival traveltimes of each pair of forward and reverse traces and produces a measure of the depth to the refracting interface in units of time which is equivalent to the time‐depth function of the generalized reciprocal method (GRM). Convolution also multiplies the amplitudes of first‐arrival signals. To a good approximation, this operation compensates for the large effects of geometrical spreading, with the result that the convolved amplitude is essentially proportional to the square of the head coefficient. The signal‐to‐noise (S/N) ratios of the RCS show much less variation than those on the original shot records. The head coefficient is approximately proportional to the ratio of the specific acoustic impedances in the upper layer and in the refractor. The convolved amplitudes or the equivalent shot amplitude products can be useful in resolving ambiguities in the determination of wave speeds. The RCS can also include a separation between each pair of forward and reverse traces in order to accommodate the offset distance in a manner similar to the XY spacing of the GRM. The use of finite XY values improves the resolution of lateral variations in both amplitudes and time‐depths. The use of amplitudes with 3D data effectively improves the spatial resolution of wave speeds by almost an order of magnitude. Amplitudes provide a measure of refractor wave speeds at each detector, whereas the analysis of traveltimes provides a measure over several detectors, commonly a minimum of six. The ratio of amplitudes obtained with different shot azimuths provides a detailed qualitative measure of azimuthal anisotropy and, in turn, of rock fabric. The RCS facilitates the stacking of refraction data in a manner similar to the common‐midpoint methods of reflection seismology. It can significantly improve S/N ratios.Most of the data processing with the RCS, as with the GRM, is carried out in the time domain, rather than in the depth domain. This is a significant advantage because the realities of undetected layers, incomplete sampling of the detected layers and inappropriate sampling in the horizontal rather than the vertical direction result in traveltime data that are neither a complete, an accurate nor a representative portrayal of the wave‐speed stratification. The RCS facilitates the advancement of shallow refraction seismology through the application of current seismic reflection acquisition, processing and interpretation technology.     

Numerical modeling of Gaussian beam propagation and diffraction in inhomogeneous media based on the complex eikonal equation

Gaussian beam is an important complex geometrical optical technology for modeling seismic wave propagation and diffraction in the subsurface with complex geological structure. Current methods for Gaussian beam modeling rely on the dynamic ray tracing and the evanescent wave tracking. However, the dynamic ray tracing method is based on the paraxial ray approximation and the evanescent wave tracking method cannot describe strongly evanescent fields. This leads to inaccuracy of the computed wave fields in the region with a strong inhomogeneous medium. To address this problem, we compute Gaussian beam wave fields using the complex phase by directly solving the complex eikonal equation. In this method, the fast marching method, which is widely used for phase calculation, is combined with Gauss–Newton optimization algorithm to obtain the complex phase at the regular grid points. The main theoretical challenge in combination of this method with Gaussian beam modeling is to address the irregular boundary near the curved central ray. To cope with this challenge, we present the non-uniform finite difference operator and a modified fast marching method. The numerical results confirm the proposed approach.     

INMOD—TWO DIMENSIONAL INVERSE MODELING ALGORITHM BASED ON RAY THEORY*

Two dimensional inverse modeling, a process to be applied after standard processing and interpretation, uses interfaces picked by the user. These interfaces are transformed into an approximate subsurface model. The subsurface model is represented by curved interfaces and interval velocities. The interfaces have to be unique functions of the line coordinate. Otherwise they may be arbitrarily curved and may begin or terminate anywhere along the section, e.g., at faults, pinchouts, salt domes and the like. Interval velocities may vary laterally along the section. The inverse modeling algorithm then modifies the model until traveltimes calculated from this model match the traveltimes observed as closely as possible in a least squares sense. The traveltimes corresponding to the model are obtained through ray tracing taking exact account of refraction. The traveltimes observed are the arrival times of single impulses before stacking contributing to the interfaces. These traveltimes are provided by ANAKON, a continuous interface analysis system. The comparison of INMOD results with those of well measurements and those of classical interval velocity computation from seismic data shows the accuracy of the method. Deviations of INMOD derived interface depths are within 2% of well data.     

Comparison of six different methods for calculating traveltimes

We study six different methods for the calculation of seismic traveltimes. All methods yield traveltimes at all points of a regular grid.
The methods examined comprise three different variants of finite-difference (FD) eikonal solvers, the graph method, wavefront construction and a combined FD and Runge–Kutta method.
The main points of investigation are computational time, accuracy and memory requirements. We took measures to obtain a high level of both general validity and clear understanding of the results. We used a profiling program to be able to measure the time that the actual core algorithm needs, thus avoiding any overhead of highly system-dependent in-/output operations.
The comparison shows that no single method is the most appropriate but that the choice depends on the task to be fulfilled. The FD eikonal solver that uses expanding squares proves to be best suited for models which are not too complicated because it offers the best compromise between speed and accuracy, whereas wavefront construction should be applied to complex media because of its superior reliability which then justifies the much higher computational times.     

Traveltime computation by wavefront-orientated ray tracing

For multivalued traveltime computation on dense grids, we propose a wavefront‐orientated ray‐tracing (WRT) technique. At the source, we start with a few rays which are propagated stepwise through a smooth two‐dimensional (2D) velocity model. The ray field is examined at wavefronts and a new ray might be inserted between two adjacent rays if one of the following criteria is satisfied: (1) the distance between the two rays is larger than a predefined threshold; (2) the difference in wavefront curvature between the rays is larger than a predefined threshold; (3) the adjacent rays intersect. The last two criteria may lead to oversampling by rays in caustic regions. To avoid this oversampling, we do not insert a ray if the distance between adjacent rays is smaller than a predefined threshold. We insert the new ray by tracing it from the source. This approach leads to an improved accuracy compared with the insertion of a new ray by interpolation, which is the method usually applied in wavefront construction. The traveltimes computed along the rays are used for the estimation of traveltimes on a rectangular grid. This estimation is carried out within a region bounded by adjacent wavefronts and rays. As for the insertion criterion, we consider the wavefront curvature and extrapolate the traveltimes, up to the second order, from the intersection points between rays and wavefronts to a gridpoint. The extrapolated values are weighted with respect to the distances to wavefronts and rays. Because dynamic ray tracing is not applied, we approximate the wavefront curvature at a given point using the slowness vector at this point and an adjacent point on the same wavefront. The efficiency of the WRT technique is strongly dependent on the input parameters which control the wavefront and ray densities. On the basis of traveltimes computed in a smoothed Marmousi model, we analyse these dependences and suggest some rules for a correct choice of input parameters. With suitable input parameters, the WRT technique allows an accurate traveltime computation using a small number of rays and wavefronts.     

跨孔观测地震数据的速度重建

利用矢量射线追踪正演模拟技术 ,计算地震波传播的路径及走时 ,进而利用射线走时及路径的内插 ,发展了弯曲射线迭代反演技术 .该方法可用来重建井间地层的速度图像 .基于所发展的方法 ,我们对两种较为复杂的典型地质模型进行了井间速度重建 .结果表明该方法是一项快速、高精度的跨孔数据速度重建技术     

地震波走时和射线的有限差分计算

以往都是采用射线追踪的方法计算地震波的走时和射线，但是当速度模型复杂时这种方法存在一些问题。本文提出另一种计算地震波走时和射线的方法。该方法从程函方程出发，利用互换原理和Ｆｅｒｍａｔ原理计算出各种波的到时和射线。解决了射线追踪方法存在的问题。为地震波走时和射线的计算以及地震波走时反演开辟了一条新途径。     

TI介质局部角度域射线追踪与叠前深度偏移成像

研究与实践表明,对于长偏移距、宽方位地震数据,忽略各向异性会明显降低成像质量,影响储层预测与描述的精度.针对典型的横向各向同性(TI)介质,本文面向深度域构造成像与偏移速度分析的需要,研究基于射线理论的局部角度域叠前深度偏移成像方法.它除了像传统Kirchhoff叠前深度偏移那样输出成像剖面和炮检距域的共成像点道集,还遵循地震波在成像点处的局部方向特征、基于扩展的脉冲响应叠加原理获得入射角度域和照明角度域的成像结果.为了方便快捷地实现TI介质射线走时与局部角度信息的计算,文中讨论和对比了两种改进的射线追踪方法:一种采用从经典各向异性介质射线方程演变而来的由相速度表征的简便形式;另一种采用由对称轴垂直的TI(即VTI)介质声学近似qP波波动方程推导出来的射线方程.文中通过坐标旋转将其扩展到了对称轴倾斜的TI(即TTI)介质.国际上通用的理论模型合成数据偏移试验表明,本文方法既适用于复杂构造成像,又可为TI介质深度域偏移速度分析与模型建立提供高效的偏移引擎.     

Eigenrays in 3D heterogeneous anisotropic media,Part I: Kinematics

We present a new ray bending approach, referred to as the Eigenray method, for solving two‐point boundary‐value kinematic and dynamic ray tracing problems in 3D smooth heterogeneous general anisotropic elastic media. The proposed Eigenray method is aimed to provide reliable stationary ray path solutions and their dynamic characteristics, in cases where conventional initial‐value ray shooting methods, followed by numerical convergence techniques, become challenging. The kinematic ray bending solution corresponds to the vanishing first traveltime variation, leading to a stationary path between two fixed endpoints (Fermat's principle), and is governed by the nonlinear second‐order Euler–Lagrange equation. The solution is based on a finite‐element approach, applying the weak formulation that reduces the Euler–Lagrange second‐order ordinary differential equation to the first‐order weighted‐residual nonlinear algebraic equation set. For the kinematic finite‐element problem, the degrees of freedom are discretized nodal locations and directions along the ray trajectory, where the values between the nodes are accurately and naturally defined with the Hermite polynomial interpolation. The target function to be minimized includes two essential penalty (constraint) terms, related to the distribution of the nodes along the path and to the normalization of the ray direction. We distinguish between two target functions triggered by the two possible types of stationary rays: a minimum traveltime and a saddle‐point solution (due to caustics). The minimization process involves the computation of the global (all‐node) traveltime gradient vector and the traveltime Hessian matrix. The traveltime Hessian is used for the minimization process, analysing the type of the stationary ray, and for computing the geometric spreading of the entire resolved stationary ray path. The latter, however, is not a replacement for the dynamic ray tracing solution, since it does not deliver the geometric spreading for intermediate points along the ray, nor the analysis of caustics. Finally, we demonstrate the efficiency and accuracy of the proposed method along three canonical examples.     

一种改进的地震反射层析成像方法

针对复杂介质的地震反射走时层析成像存在数据拾取困难问题，本文提出了一种新的地震反射层析成像速度模型建立方法，该方法用速度和地震射线走时描述模型，用地震反射波走时、地震波在源点和接收点处的传播方向信息反演模型.为提高反演的稳定性和计算效率，引入了Hamilton函数描述射线，在相空间计算反演所需的射线路径和目标函数对模型参数的导数，对理论模型和实际地震资料进行了试算，试算表明该方法对复杂介质具有较强的适应能力.     

一种改进的线性走时插值射线追踪算法

线性走时插值法（LTI）在走时的计算中,由于射线方向考虑不全,计算得到的节点走时不一定最小,导致追踪的射线路径无法满足最小走时.针对这一问题,本文提出了一种改进的射线追踪算法,通过采用多方向的循环计算,得到所有计算节点的最小走时,使追踪到的射线路径能真正满足最小走时,以确保射线追踪的精度.模拟实验结果表明,在介质速度变化剧烈的结构中,该算法与传统的LTI算法相比,有效地提高了射线追踪的精度.     

Moveout approximation for vertical seismic profile geometry in a 2D model with anisotropic layers

I introduce a new explicit form of vertical seismic profile (VSP) traveltime approximation for a 2D model with non‐horizontal boundaries and anisotropic layers. The goal of the new approximation is to dramatically decrease the cost of time calculations by reducing the number of calculated rays in a complex multi‐layered anisotropic model for VSP walkaway data with many sources. This traveltime approximation extends the generalized moveout approximation proposed by Fomel and Stovas. The new equation is designed for borehole seismic geometry where the receivers are placed in a well while the sources are on the surface. For this, the time‐offset function is presented as a sum of odd and even functions. Coefficients in this approximation are determined by calculating the traveltime and its first‐ and second‐order derivatives at five specific rays. Once these coefficients are determined, the traveltimes at other rays are calculated by this approximation. Testing this new approximation on a 2D anisotropic model with dipping boundaries shows its very high accuracy for offsets three times the reflector depths. The new approximation can be used for 2D anisotropic models with tilted symmetry axes for practical VSP geometry calculations. The new explicit approximation eliminates the need of massive ray tracing in a complicated velocity model for multi‐source VSP surveys. This method is designed not for NMO correction but for replacing conventional ray tracing for time calculations.     

多步法快速射线追踪与圆台型高精度走时外插计算

为更好地适应复杂构造的地震偏移成像,本文提出了一套快速射线追踪算法和一种高精度的走时外插计算方法.采用线性多步法的预测-校正公式求解射线追踪方程组,与传统的四阶Runge-Kutta法相比,提高了计算效率.在网格节点上的走时计算中,应用一种基于圆台的外插方法,该方法以射线的方向为轴确定圆台,将轴上的走时外插到圆台内的网格节点上.与传统的矩形体外插方法相比,圆台走时外插方法提高了计算精度,且具有更好的稳定性.另外,该方法利用稀疏分布的射线即可获得高精度的走时表,节省计算量,对复杂构造的偏移成像非常有利,尤其是三维偏移.最后通过逆散射偏移成像算例,验证了算法的有效性和适用性.     

地震走时CT波前追踪

把走时CT结果作为地震衍射CT(DCT)的背景场可以改善DCT之成像质量。本文由Radon变换、泛函分析变分原理和微分几何导出程函方程后,采取有限差分法求解,实行波前追踪。通过对反射地震勘探和VSP 结构的源——接收系统编程实际运算,表明该波前追踪程序输入简单,适合于多层介质模型,各网格点的走时能迅速确定。     

A hybrid fast algorithm for first arrivals tomography

A hybrid algorithm, combining Monte-Carlo optimization with simultaneous iterative reconstructive technique (SIRT) tomography, is used to invert first arrival traveltimes from seismic data for building a velocity model. Stochastic algorithms may localize a point around the global minimum of the misfit function but are not suitable for identifying the precise solution. On the other hand, a tomographic model reconstruction, based on a local linearization, will only be successful if an initial model already close to the best solution is available. To overcome these problems, in the method proposed here, a first model obtained using a classical Monte Carlo-based optimization is used as a good initial guess for starting the local search with the SIRT tomographic reconstruction. In the forward problem, the first-break times are calculated by solving the eikonal equation through a velocity model with a fast finite-difference method instead of the traditional slow ray-tracing technique. In addition, for the SIRT tomography the seismic energy from sources to receivers is propagated by applying a fast Fresnel volume approach which when combined with turning rays can handle models with both positive and negative velocity gradients. The performance of this two-step optimization scheme has been tested on synthetic and field data for building a geologically plausible velocity model.This is an efficient and fast search mechanism, which permits insertion of geophysical, geological and geodynamic a priori constraints into the grid model and ray path is completed avoided. Extension of the technique to 3D data and also to the solution of 'static correction' problems is easily feasible.