中文摘要：

初至波走时层析是获取近地表速度结构的一种常用方法.随着采集技术的不断发展,可使用的数据量迅速增多,传统的基于射线追踪和解方程组的地震走时层析成像方法面临着内存占用大、方程求解不稳定等问题.为了解决这些问题,本文基于前人在波形反演研究中提出的一种改进的散射积分算法,提出了一种预条件最速下降法初至波走时层析.该方法无需存储核函数矩阵与Hessian矩阵即可方便地实现目标函数梯度的计算与预条件,且该方法计算效率高、求解稳定、易于并行.数值实验结果表明,该方法可以获得与传统方法精度相当的反演结果,但所占用的内存大幅减小.

英文摘要：

First arrival traveltime tomography is a commonly used method at present to obtain the near-surface velocity structure.With the development of acquisition technology,the data quantity is growing rapidly.Thus,traditional ray-based equations-solving method faces the challenges of memory consuming and instability.To solve these problems,an improved scattering-integral（SI）algorithm proposed in a former study on full-waveform inversion（FWI）is employed here to develop a new preconditioned steepest-descent ray-based first arrival traveltime tomography.This method can compute the gradient of the objective function and implement the precondition conveniently without storing the sensitivity kernel matrix or Hessian matrix in advance.In addition,the inversion process is efficient,stable and easy to be paralleled.Numerical experiments show that this method can achieve accurate inversion results comparable to traditional methods,while the memory requirement is significantly reduced.