Reflection tomography is widely used to build velocity models for seismic depth imaging, but several factors can limit its effectiveness, particularly in resolving complexities in the near surface. A 3D data set acquired in a mountainous area of western China exhibited several such factors, including low fold of coverage in the shallow data that made it difficult to verify and adjust velocities based on measuring the flatness of common-imaging-point (CIP) gathers. The data set also presented challenges related to poor signal-to-noise ratio (S/N) and dramatic lateral velocity variation in the shallow section. The diving-wave tomography method, combined with precise first-arrival picking, provided an enhanced solution for deriving the shallow velocity field in the initial depth-migration velocity model. This was part of the innovative and enhanced velocity model-building workflow that was applied to accurately image the complex subsurface geology of this challenging data set. The workflow comprised four key steps: data conditioning, diving-wave tomography, reflection tomography, and prestack Kirchhoff depth migration. This project was the first application of diving-wave tomography combined with reflection tomography to land seismic data in China. Its demonstrated success provides a robust and effective velocity model-building solution for application in other complex depth-imaging land seismic projects in China and elsewhere.