Abstract: Seismic anisotropy is one of the important indicators reflecting internal medium character of the Earth.The commonly used shear wave splitting method and two-dimensional(2D) tomography of surface-wave azimuthal anisotropy are difficult to accurately reflect the change in anisotropy with depth.Converting the period-dependent regionalized surface-wave azimuthal anisotropy into depth-dependent one-dimensional(1D) S-wave velocity azimuthal anisotropy is one way to make up for the lack of depth information.The existing researches on three-dimensional(3D) S-wave velocity anisotropy are normally implemented by a two-step method including period-by-period 2D tomography of surface-wave azimuthal anisotropy and grid-by-grid 1D inversions for S-wave velocity anisotropy.This stepwise method is neither conducive to the use of 3D a priori constraints,nor conducive to the direct evaluation of the inverted 3D model by checking misfit of original observational data.Therefore,we developed a method on 3D tomography of S-wave velocity azimuthal anisotropy by using surface-wave dispersion curves and implemented the relative forward and inversion programs.In order to check the effectiveness of the methods and procedures,we conducted synthetic tests for regularly distributed checkerboard models.Test results show that:this method can well recover 3D structural information such as isotropic velocity anomalies,relative anisotropic intensity,and fast wave direction;the inverted model significantly improves the fit and reduces the misfit to the observation data compared to the reference model.However,given an isotropic synthetic model,the 3D anisotropic inversion can result in<0.5%of false anisotropy in homogenous areas,and the false anisotropy increases for strongly heterogeneous areas,reaches up to 3.5%at shallow depths.So,in practical studies,interpretation on anisotropy in strongly heterogenous area(especially at shallow depths) must be cautiously treated.